NDA_QMC.jl

Open the notebook NDA_Notebook.ipynb to see what to do.

This is the repository for the SN + NDA + QMC + GMRES stuff

This repo is part of the CEMeNT project

If you're not a CEMeNTer, go away.

Research Issue: How to get NDA to work with QMC

One problem, probably not the only one, is that d phi/dx is not correct. Sam and I have tried a couple different ways without success. The new notebook QMC-NDA-Woes.ipynb explains where we are so far. This mess will get merged with the real notebook when we get our problems figured out.

Contents

• Instructions
• Dependencies get these packages first
• Problems?

What to do.

• Clone this repo
• Install the dependencies
• Fire up IJulia and cd to the directory where you put the repo
• Open the notebook NDA_Notebook.ipynb
• Do a run all this will run nda_note.jl which gets the dependencies organized and puts the right things in your load path
• You are now ready to mess with the sovlers. The codes in the /src directory are pretty well documented and ready for you to play with.

The transport solves

I'm duplicting the results from

R.D.M. Garcia and C.E. Siewert. Radiative transfer in finite inhomogeneous plane-parallel atmospheres. J. Quant. Spectrosc. Radiat. Transfer, 27:141-148, 1982.

I run two classes of sovlers so far

• Plain-vanilla source iteration and GMRES accelerated source iteration
• NDA with Picard iteration and NDA with Newton-GMRES
• QMC with Picard and GRMES. GMRES works great because QMC is deterministic.

QMC with NDA next on the list.

Dependencies:

• SIAMFANLEquations.jl

  • You need v 0.3.0 or higher. Get it from the pkg REPL with

  add SIAMFANLEquations

• FastGaussQuadrature

• LaTeXStrings

• PyPlot

• SuiteSparse

• SpecialFunctions

• LinearAlgebra

• Sobol

• JLD2

• Random

Problems

If you're CEMeNTer, you know how to find me. Otherwise ... go away.