Code for Conway's Ph.D., for Mathematica Version 12.
For the thesis, see yawnoc/phd-thesis.
WARNING.
This repository contains upwards of 200 MB in objects
(with a 500 MB working tree),
mostly compressed NDSolve`FEM` solutions,
which took a long time to compute.
In hindsight, it would have been better to store these files separately,
so as to not bloat this repository.
Including the large output files was a stupid way of ensuring identical copies on a large number of computers. While such a setup was reasonable using a USB stick as a shared Git remote, fetching 200 MB per clone over the internet is not a good thing.
To save bandwidth, consider cloning yawnoc/phd-code-no-output instead, which is a debloated version of this repository with no output files. Only download the large output files individually as needed.
LaTeX fonts in the figures were produced without Szabolcs's MaTeX.
While MaTeX is an excellent package outputting vector text labels,
one needs a TeX system and Ghostscript.
Instead I used a simple character mapping
along with Mathematica's own string formatting capabilities
(see the functions LatinModernFont, LatinModernLabelStyle,
LatinModernFontStyle, and LaTeXStyle in Conway.wl).
This approach only requires the installation of the following fonts:
-
"Latin Modern Math"
latinmodern-math.otf
http://www.gust.org.pl/projects/e-foundry/lm-math/download -
"LM Roman 10"
lmroman10-*.otf
http://www.gust.org.pl/projects/e-foundry/latin-modern/download
On Windows there are some issues with spacing around Greek characters due to a bug ... the solution is to use Linux instead.
Nevertheless, Linux appears to have a bug where capital Greek letters fail to be italicised when exporting to PDF. Therefore use Windows for radiation-conduction-bvp.pdf, capillary-contact-angle.pdf, and helmholtz-wedge-domain.pdf.
Using Mathematica to do applied maths is an exploratory process.
Each section of the included notebooks contains explorations
which I thought useful at the time of writing.
Some are Manipulate visualisations,
some are static plots which I have exported as reminders for myself,
and some are complete dead ends.
Don't simply run all the code in a notebook.
This will not be very illuminating,
not to mention time-consuming since you will end up
doing computations which needn't be redone
(assuming you aren't using the debloated version).
Usually you want to run only the initialisation section.
This is the section I have called
"Initialisation section (always run this first)";
select its cell bracket and evaluate.
Do NOT use "Evaluate Initialization Cells".
(All cells in a .wl file are automatically styled as Code cells,
which are Initialization Cells by default.)
Timings referred to in the comments are for an
Intel(R) Core(TM) i7-7700 CPU @ 3.60 GHz (8 GB RAM)
with a Benchmark[] score of about 2.4.
This repository is licensed under "MIT No Attribution" (MIT-0), see LICENSE.