Schedule of meetings for semester

Schedule for this week

Research review

I've implemented GNU portable threads and have done more scaling tests on bluehive (over ethernet). I've been combing through the data making figures for the paper's results section.

I am implementing the viscosity and resistivity :
The resistivity solver is similar to the talk I gave last week, which implements a scalar resistivity that depends on the local temperature. We can be more sophisticated by allowing the resistivity to be a tensor: remember, the field diffusivity across the flow velocity and along the flow velocity is different by a factor of 2. But I have not discovered yet any MHD code fully implement this complexity. The viscosity is calculated similarly. Currently, the scalar viscosity is considered. We can implement a viscosity tensor later, but it will also be slightly more complicated.
Both the viscosity and resistivity modules are currently finished, which computes the source term induced by these two physics processes at cell centers and face centers respectively. They also calculate a preferred subcycling time based on the CFL number. Although I have not decided how to implement the subcycling since as we discussed before, the energy flow induced by the resistive diffusion and viscous diffusion are likely ignored if we do subcyclilng only in the operator splitted diffusion equation. If we just compare the hydro time step with these time steps, we can get the wanted temporal resolution for energy evolution, but that will mean a slower code.
The next step is to develop a coupling method to integrate these source terms into our MHD solver. I have discussed a bit of this with Jonathan last week. Hopefully we can get some thing running this week.
Also finished editing the paper. It is ready for submission now.

Since I won't be at the meeting tomorrow on account of studying for the prelim, here's what my schedule looks like this semester…

Galactic Dynamics - TR 2:00 - 3:15 Intro to Plasma - TR 3:25 - 4:40 Hydro Stability - MW 4:50 - 6:05 (If I do take this course it will only be an audit, so consider this time slot as open if needed)

1) Monday 4 pm: The usual research review 2) Tuesday 4pm: A special new users group meeting where WE ALL are new users and will do a general code review. Jonathan will present an overview of the AMR methodology and the current code capabilities and we will then run through an overview of what the code can can not do. 3) Thursday morning: (Optional for Most except but Eddie and Jason should attend if possible): We will be running experiments at the LLE on Wednesday and on thursday morning there will be a meeting to review the shots and more importantly discuss future directions. More details will follow

• The initialization of the Bonner Ebert sphere looks as expected.

• Decreasing the thermal energy in the grid leads to collapse.

• Reading the Federrath et al. paper on sink particle implementation now. One of the tests they did was the collapse of a singular isothermal sphere (this is cool). They found the collapse to match Shu's solution, and the accretion rate of material onto the sink also matched Shu's solution for M-dot.

• Would like to finish up this summer's numerical study of the collapse of the BE sphere by relaxing the sinks criteria so as to form a sink during the collapse.

New AMR tests. I am trying to implement a few lines of code in ProblemSetErrFlags routine in problem.f90 to refine only within some center region of my Bonner Ebert sphere. Here is what I found:

​http://www.pas.rochester.edu/~erica/mesh_test1.html

I put some useful scripts on the wiki for running multiple jobs with various parameters on different numbers of processors.

Scaling Scripts

Both Hydro runs are almost completed on bluehive. The magnetized runs were dying (see comment:58:ticket:121) but I've reduced the default message size and resubmitted the jobs. I could also move these over to bluehive once the scaling tests are completed…

Parsing data from scaling tests. Collected 20+ statistics from 2800 runs… Preliminary plots are attached to ScramblerPaper

• Visit use on bluehive up on tutorial page
• Bonner collapse update - ​http://www.pas.rochester.edu/~erica/Collapse.html

Still studying for the prelim like a madman. In regards to research, I need to figure out how to turn on uniform gravity for my RT module. Shule pointed me in the right direction last week, but I'm not sure if it's working yet…update on this to come later this week.

CRL 618
I made a movie of the clump so that people can see the formations on the head of the clump. ​Movie
I'm now focusing all 64 of my Bluehive processors on getting the jet runs done, which are currently going painfully slow.

Radiative Shocks
I've completed runs of the latest module for alpha = 0, 1/3, ½, 1 and the results look good. Movies, temperature lineouts, and plots of shock front vs time are in the process of being made.

REU talk
Here is the pdf of my REU talk on 8/5: ​Talk

Just reading over the Foster & Chevalier

​http://adsabs.harvard.edu/abs/1993ApJ...416..303F

The collapse of a Bonner-Ebert Sphere appears fairly complex with solutions approaching the Larson-Penston (1969!) solution or the Shu 1977 solutions or the Hunter 1977 solutions depending on intiial conditions in the BE sphere.

I wanted to make sure that the bonner ebert sphere would behave as expected under conditions of extreme gravitational instability corresponding to values of the density contrast >> 14.1. Under such conditions, I expected the sphere to collapse to very high densities within a few free fall times. This however, did not happen. What seemed to have happened is concerning. It looks like there is a global mass loss over the grid. Please see the movies in the following link. This sphere has a density contrast ~ 200. I looked again at a sphere that should have been unstable at about rho_o/rho_c ~ 20-30, but was not. This global dip in density also occurred there.

​http://www.pas.rochester.edu/~erica/bonnerWeirdness.html

Also — NO sinks were formed ..

My calculation of mass and mBe:

-volume of ambient = volume of box - volume of sphere

!= 64 - 4/3 Pi

!=60

-mass of ambient = rho_ambient * volume ambient

!= 0.0004 * 60

!= 0.024 (mass - computational units)

-mass of sphere = query result for 'weighted variable sum' of rho (of entire box) - mass of ambient

!= 0.466 (mass - computational units)

To get into cgs, multiply by (rscale/lscale³)?

This gives: 1.57E-73..

I then compared this to mBe:

mBe = [1.18 * (aT)⁴] / [P^½ * G^3/2] My simulation gave a 9690 cm/s sound speed (aT) and a pressure of 2.87E-19.

These answers mismatch by an enormous factor, some 50 decades…

8/10/11

Today I set all qtolerances to be a minimum of 1d-16 after observing the slight asymmetry of the mesh was elliminated when switching from 0.1 to 1e-16:

​http://www.pas.rochester.edu/~erica/vvec2.html

I also wanted to maximize the refinement on my grid.

Next I tried playing with the refine variable option for density to see how it would decrease the refinement in my grid. I ran simulations of 1, .1, .0001, 1e-64.

refVar = 1 expected no change — this was right. But as I decreased the value of refVar, I didn't see an obvious decrease in refinement… Once I got to 1e-64, the grid had lost all refinement except for the center 8 cells that refined to level 1. (I was running level = 2).

I also tried refVar = 0d0, but can't seem to get rid of that center region of refinement like I expected.

I tried adjusting the fill ratios from constant 0.9 to 0.3. This created a square region of refinement — Still can't get center region of refinement to go away with refVar = 0. Ditto for fill ratio of 1..

​http://www.pas.rochester.edu/~erica/vvec2.html

I looked in the clump and ambient object modules, nothing seems to be overriding the setflags routines there…

8/9/11

Working on understanding the amr controls in astrobear.

​http://www.pas.rochester.edu/~erica/refVar.html

Here is a link to the paper we discussed about a new method for cooling

​http://adsabs.harvard.edu/abs/2009ApJS..181..391T

And here is a link from the guy who told me about it Allard Jan van Marle on AMR and cooling

​http://esoads.eso.org/abs/2010arXiv1011.2610V

I am going to try and attach my latest PDF daily progress report on doing gas-filled targets.

​Dink

I've really been focusing on studying for the prelim for the past couple of weeks, so I have not done much in terms of research. Perhaps sometime this week Shule can show me how he calculated growth rates for the RT Instability. Also, once Martin finishes the scripts for the testing suite, I will be able to do more towards that project. However, the prelim will remain my top priority for the month of August.

The resistive MHD is now in the code. I've added stuff to the project page on wiki. ​https://clover.pas.rochester.edu/trac/astrobear/wiki/AstroBearProjects/resistiveMHD

There are simulations and some discussion on it. This should be the basis of our upcoming magnetized cloud project, together with the force-free field stuff.

Confirmed truelove collapse problem still works correctly

Uniform Collapse page and movie

The particle wanders and there are a few things I want to try to determine the cause…

• Switch from using \frac{\partial \phi}{\partial t} from previous coarse advance, to the current coarse advance. This prevents level threading but should still give good performance.

• And do a resolution study to see how it affects the outcome.

• And check the code to make sure the accretion forces and the gas forces on the particle have the proper signs.

(Just finished a level 4 run with no significant drifting… Might be a restart issue)

• New Users Meeting Page, ​https://clover.pas.rochester.edu/trac/astrobear/wiki/u/NewUsers
• BE project page, ​https://clover.pas.rochester.edu/trac/astrobear/wiki/BonnerEbert
• Research blog page? ​http://astrobearcub.blogspot.com/
• Library page, ​https://clover.pas.rochester.edu/trac/astrobear/wiki/u/EricasLibrary
• Can use visit on bluehive. Added some notes on how to do so on the visit tutorials page. Should use Bluehive's compute nodes instead of the login nodes…

CRL 618 simulations are running in Bluehive with added rings/ripples in the ambient medium. Bluehive queue has been busy lately, but so far:
Clump: ​http://www.pas.rochester.edu/~blin/visit0000.png We had to lower the refinement variable factor in physics.data for the clumps because the chombo file sizes we were getting were huge ~2gb.
Jet: ​http://www.pas.rochester.edu/~blin/visit0005.png Martin has a 0amr version of the jet which is complete.

Not much update on radiative instability work due to focusing on REU paperworks and talk. Radiative inst. module tests will be the main focus this week while waiting for CRL 618 runs.

​http://www.pas.rochester.edu/~erica/vvec.html

I am running into some odd behavior with my bonner ebert simulations. The simulations run smoothly for about 3 crossing times. During this time there is a density increase of about 500x and the mass is squeezed into a tiny region at the grid center. I have noticed that the mass here is unevenly distributed (I.e., more so along one dimension than the others). Then the entire clump expands, fast, but in a squashed geometry. It seems like as the simulation progresses, however, the clump tends again toward a spherical geometry.

I am curious why this squashing along one dimension is happening. In earlier simulations, I saw distortion caused by pressure waves running off the sides of the cube and interacting with the center sphere. There, however, the geometry remained symmetric, as it should… I am dubious as to why I am seeing this asymmetry appear.. Perhaps the volume into which the material is collapsing is smaller than the grid can resolve, causing some wonky numerical behavior? I noticed a diamond shape occuring at the center of the sphere in the slices of density contours, leading me to believe this could be a matter of resolution..But alas, why would one of the squares in the diamond contain more mass than the others?? Why would this mass not be symmetrically distributed?

ALSO — I noticed that running AMR with the same effective resolution took ~5x longer…

I am running a 63x63x63 job now, with 3 levels AMR..