Conference and UCSD Summer School

In June, I went to my first physics conference: LaserNetUS. I got to present my poster on the Fuchs ML work I have been doing. I got to see some important people in the field including Blagoje Djordevic and Scott Wilks and even got to meet Brendan Loughran from the Bayesian Optimization Paper.

On another note, the conference room was FREEZING and I was shivering the whole time. I thought it was really cool how they signaled people to walk back inside using this type of soothing chime, rather than some large announcement.

During the last two weeks of July, I attended the High Energy Density Science Summer School at the University of California at San Diego! There were lots of talks by well-known researchers in the field. I particularly liked the talk from Burkhard Militzer on theoretical aspects in planetary science. He talked about how the interior of Jupiter is still under contention and that there might be some liquid component of metallic hydrogen. Additionally, he talked about how the rings of saturn may have been formed from tearing apart one of its moons: which would change perspectives on just how old Saturn’s rings are.

There were also some good talks on ICF and we even got to visit General Atomics and view their tokamak and target fabrication facility! Additionally, there was a machine learning workshop led by Pat Knapp and Raspberry Simpson. I have pdfs of all the talks, so that will all be useful as a resource to me in the future.

Fuchs ML Work Paper

We ended up finishing up the work on the Fuchs Model Machine Learning Paper and uploaded it to arXiv! We made minor modifications since last post because I forgot to properly re-instantiate the Neural Network model each time (whoops) and I capped the number of GP iterations to 50 to avoid numerical issues. I plan to write up a more formal post on this later.

Ricky’s Energy Conservation Paper

Last year, I did a bunch of work on energy conservation in PIC sims. I found out that the Arber et. al. paper empirical fit to the amount of heating (derived from Hockney) didn’t do a great job across different resolutions (e.g. debye length to cell size ratio). This is further complicated when the ion species aren’t just protons and can have multiple levels of ionization. To summarize, the Arber formula is

\begin{equation} \frac{dT_{eV}}{dt_{ps}} = \alpha_H \frac{n_{23}^{3/2} \Delta x_{nm}^2}{n_{ppc}} \end{equation}

which shows an inverse relationship with the number of particles per cell. However, as the plasma heats up, ionization can occur which will shed more electrons into the system. If the number of ion particles per cell is too small, the electron macroparticles shed by the ions will have a lot of charge and contribute more to the electric field fluctuations. This in turn will cause more heating. Additionally, a higher ionization state should have more heating as well. Putting this together, we propose the following modification:

\begin{equation} \frac{dT_{eV}}{dt_{ps}} \sim \alpha_H n_{23}^{3/2} \Delta x_{nm}^2 \left(\frac{1}{n_{eppc}} + \frac{Z_O^* - 1}{n_{Oppc}} + \frac{Z_C^* - 1}{n_{Cppc}}\right) \end{equation}

where $Z_O$ and $Z_C$ are effective ionization states of oxygen and carbon. Ricky assumes that all ion species are initially singly ionized and further ionization will occur as the target heats up (see ionization energies here ). This seems to explain some amount of variation of the data

Here is the comparison if we allow a free parameter as a heating constant out front:

and here is the comparison if we don’t allow this free parameter:

Journal Club

Lately, we started a Journal Club where we discuss OSU dissertations, candidacy exams, lecture notes, and research papers. It has been occuring on thursdays at 12:30AM.

Things to Do

• Run an EPOCH simulation and try to confirm some of Ricky’s Results
• Learn how to use VisIt and how to extract info from .sdf files
• Modify Hockney Integral and see what changes