New Post is up.

[mattman]

New Post: "We have to Try"

The blog has past 70,000 views. Crazy. Also it is getting allot of traffic on Google+. It got noticed by the STEM community, Hackday and Science 2.0 blog. Thanks to Dan Tibbets, who helped provide feedback.

===

http://thepolywellblog.blogspot.com/201 ... o-try.html

====
Summary:

This post covers work from Convergent Scientific Incorporated. It has four parts: experimental work, modeling, talk highlights and a conclusion. Experiment details were imperfect. CSI trapped electrons for 20 seconds, using: model one. It is assume this is a wire shaped into a diamond - 14 cm a side with 1,500 amps and held at +500 volts. This was within a ~0.6 pascal vacuum with four emitters at each corner. Model one was cooled with a chilling system. The emitter voltage was varied from -500 to -9,500 volts and a probe measured the resulting trapping. Results were questioned because the emitters remained on the whole test. The magnetic and electric fields are mapped using single wire and point charge models. The excel file is open to the public. The forces are plotted along the face path of the diamond. The resulting motion is described. The effect of sliding or rotating the emitters is explored. Talk highlights are given. These include: the impact of pressurizing, moving, shaping and forming structure within the plasma. The relevant plasma instabilities are mentioned as well as a discussion of structure. Finally a call for experiments is made, with a list of good reference material.

[asdfuogh]

Modeling is for wimps:

Theory is also for wimps:

I agree with the general statements of those parts of the post, but I think it's also important to highlight that these are all just different approaches to analyzing problems and ideas (ie. computationally, analytically, experimentally). It's just that a correct experimental approach is the ultimate validation of a model, but if you are just experimenting randomly without thinking about why you might want to look at something first... well, the Edison approach to fusion devices is probably not what we want to go for, is it?

[mattman]

Hey,

I am glad you said something. There is lots of time spent working on this stuff, and if nobody reads it, it is a waste. If it took you 20 minutes to read it, I spent hundreds of hours working on it.
===
I think there is a balance there, but ultimately people (myself included) will only trust something they can see working. It was the only way people believed in flying machines. So we need to borrow from the tech startup world: try and fail, and when you fail, fail fast.

[asdfuogh]

>try and fail, and when you fail, fail fast

Ideally, that's what we want for the fusion industry as a whole, but not for every individual.

But, there's a giant cost of entry difference between current tech start-ups and fusion start-ups. With tech startups, the science and engineering is already there, and it's often a matter of doing something better than your competition (or well enough that the competition buys out your company). With fusion startups, you have to either know investors or convince your investors that you can eventually achieve something. Not every investor cares about the long-term future, and most of them probably care about getting a return on their investment.

So, if you go and proclaim that as your motto as an individual startup.. well, I don't think that's how Tri-Alpha and General Fusion persuaded their investors to come on board. What I like about those two companies are how much they're invested in theory and computational models as a parallel component to their work. I don't think blindly charging is ever a good move in experiments, but it's even more important in fusion research to take calculated steps, especially for anything that is disseminated to the public. It's not just annoying to hear people ask me "isn't fusion always 30 years away?" but it's also damaging to research and funding to have that kind of reputation. So, I think it'd be more apt to encourage researchers to "make haste slowly" instead.

[mattman]

We basically agree on this.


Everyone is interested in trapping right now. I think soon, people going to start looking for a resonance condition. A sweet spot within the operating space. A way to run it where the fusion rate peaks and the loss rates plummet. Right now, this is an optimization problem.


To find it, you will need to do Particle In Cell simulations with dimensionless numbers. You will need a benchmark first, to ensure your code is working. Several teams have started down this path: University of Sydney, Wisconsin, Randy, Radiant Matter Research, Happy Jack, the Atomic Energy Organization of Iran, Convergent Scientific Inc, The Navy, ect...

Who will win? it is an open field. And, once the simulation works, it is only the map. To prove it, you will need to build.

====
For the public, a simple visual basic or java app would help allot. Give the public a webpage with radio buttons and variable bars. Let them see how single electrons or the cloud will change as the B-Field increases 1.5X times, or the rings are doubled. Simple tools like that, can go a long way. But once again, People will need to "see it" to believe it.

[asdfuogh]

> And, once the simulation works, it is only the map. To prove it, you will need to build.

You're right. Theory always has to work hand-in-hand with experiment. You have to benchmark with other codes, and benchmark against reality.

I don't think it would be necessary to make such visual apps for the public, to be honest, not unless you're looking for a very minor amount of funding. What really is necessary are strong results from experiments and computational models to convince people with the funding of the promise of polywell (ie. private ventures or the government). I expect that the public can't really be persuaded by a pretty program considering the horrible reputation that fusion research has in the public's mind..

[mattman]

A next possible post could be ripping apart CSI simulations talk. That is, digging in and explaining it.

In CSI's second talk they laid out their simulation approach. Mr. Baker makes reference to a paper by L. Chacon. Dr. Chacon was a graduate student under Dr. George Miley at the University of Illnois at Urbana around 2000. He did some fusor work for Miley. Since graduating Chacon has worked at Los Alamos National labs. He has published a bunch of paper simulating and exploring the physics of plasma clouds (I have 15 or so). I want to get the paper CSI is using, and explain it.


There are some interesting tie-ins with other simulation work:
========================================
1. Happy Jack, Randy and Indrek have all simulated the polywell using one code or another, their stuff can be downloaded for free.
2. There is a company in Wisconsin which has great PIC code. The Iranians used this for their 2011 paper.
3. Jeffery Kollasch at Wisconsin is using a Vlasov-Poisson model to simulate the reactor. What is this? How does it work? What are the limitations? I may email him directly with questions..
4. CSI is using CUDA. What is CUDA? Happyjack can help here. How do we use it? Ect...
5. The gold standard in Fusion is MCNP6. There is a website & tutorals looking at how to use this code. Could polywell simulations be done using this?
6. In a post from last year, I explored the idea of simulating WB6 deeply. I also looked at the history of the particle-in-cell method. I estimated how many particles one would need to simulate WB6. It would be interesting to compare that post to CSI approach.



===
Accuracy is key. Clarity is key. Jargon must become plain English. But, simplicity & accuracy is hard to get. That is why these posts can take a long time. That is why they are never perfect.

[prestonbarrows]

Vlasov-Poisson is a set of equations which can numerically approximate a plasma.

There is more or less a spectrum of modeling approaches; solving for the trajectories of individual particles on one end and using fluid-like dynamics on the other end.

Ideally, you would solve the trajectory of each ion and electron. These are very simple equations. the issue is there is an untenable number of particles to solve for in parallel. These are often called kinetic descriptions. They tend to work better for low densities or non-thermal plasma where simplifying assumptions can't be applied.

Fluid descriptions only keep track of macro-scale properties like pressure or temperature; the individual particles are ignored. This means they are much faster to calculate, but tend to include a lot of assumptions and simplifications or require empirical data.

Vlasov-Poisson is somewhere in the middle, dealing with distribution functions of particles. I don't have much direct experience with this approach.

CUDA is a codebase for GPUs developed by NVIDIA. In this context, it allows for much faster, parallel, computation as compared to a CPU.

MCNP is not applicable directly to plasma simulations. It is designed for calculating behavior of neutrons, gammas, and beta radiation as it propagates through materials. This is useful for designing fission reactors or shielding around a fusion reactor, but not the plasma itself.

[asdfuogh]

When you talk about simulation, you also have to know the corresponding physical time scale. In order to run an accurate kinetic or gyrokinetic code (or even fluid codes), you'll need to know how to simplify the physical equations into computational models that you can program into a code that will finish running before the end of the universe. The time-scale you impose on the simulation will change the physics you can possibly see (ie. you will filter out waves by choosing a time-scale to simulate).

What exactly did CSI simulate?

[tombo]

Awesome!
Thank You!

We always knew that joule heating would be a major problem.