mitchellmcm27 / tcg-ec

Clone the repo

git clone https://github.com/mitchellmcm27/tcg-ec.git
cd tcg-ec

We also need to pull in the "TCG_SLB" code, which provides convenient Python classes and scripts for working with the Stixrude & Lithgow-Bertelloni (2011, 2021) databases through TCG. One way to do this is to simply clone it in. From inside the tcg-ec directory, run the following. This will create a tcg_slb subdirectory containing the required code.

git clone https://gitlab.com/cianwilson/tcg_slb.git --depth 1

NOTE: I will probably need to update this part of the instructions before releasing the code.

Start a Docker container

Build an image using the provided Dockerfile, using -t to give the image a human-readable tag. Start an interactive container, binding the local tcg-ec directory to workspaces/tcg-ec inside the container.

docker build -t tcg-ec .
docker run -it --rm -v $PWD:/workspaces/tcg-ec tcg-ec

Alternatively, open the tcg-ec directory in VSCode and use the Dev Containers extension to automatically build and re-open the repository inside a Docker container.

Thermodynamic database and reactions

A custom thermodynamic database based on Stixrude and Lithgow-Bertelloni (2021) is included as tcg_slb/database/tcg_stx21_database.tar.gz. The scripts, source code, and data for generating this database are also provided, but doing so is not necessary.

Reactions are included as *.rxml files. Because generating the C++ code for these reactions can take several hours, pre-built binaries and Python bindings are included, which are compatible with the Docker image.

If reactions are edited and need to be re-built, you can do so as follows:

cd tcg_slb
scripts/generate_reactions_eclogite -v 21
scripts/build-reactions database/reactions/eclogitization_2024_stx21_rx.rxml

Model calculations

Model calculation scripts are in the models directory.

cd models

Four python scripts are given as follows:

• python3 parallel_pd.py generates a (T,P) pseudosection for comparing density with Perple_X results.
• python3 parallel_profile.py generates a 1-d profile through (T,P)-space for comparing phase mode with Perple_X results.
• python3 parallel_experiment2.py runs the geodynamic model of crustal thickening at the Moho.
• damkohler-fit.ipynb shows how Damkohler number is fit to empirical data.

In most cases, you can pass the name of any pre-defined composition that exists in the models/perple_x/compositions.json file. For example, python3 parallel_pd.py -c hacker_2015_md_xenolith. By default the parallel_* scripts use all available CPU cores.

Arguments can be passed as follows to customize the model runs:

CLI argument	Purpose	Default
-n [int]	Number of CPU processes to use	mp.cpu_count()
-e [float]	Scaled ending time	1
-c [string]	Bulk composition to use, by name	an array of 4 compositions
-r [string]	Reaction to use, by name	eclogitization_2024_stx21_rx
-q	Run in "quick mode" (Da ≤ 1e4)	False
-f	Force model to recalculate results	False
-p [string]	String to prefix output directory	None

Model outputs

All outputs are saved to the models/output directory. Outputs will be automatically grouped into subdirectories based on the name of the reaction and composition.

Perple_X: Adding new compositions

Perple_X outputs for equilibrium thermodynamic properties are included for several compositions defined in the models/perple_x directory. The python scripts read the text files output by Perple_X to initialize the reactive thermodynamic models.

If a composition needs to be added that does not already exist, use the following recipe:

• Add the oxide percentages to models/perple_x/compositions.json file, making sure to use the same template as the existing compositions.
• Within the perple_x directory, run julia solve_composition.jl [name] where [name] is the unique identifier, the key for the composition object in compositions.json (e.g., sammon_2021_lower_crust).
• The composition can now be used in python scripts by passing the -c [name] argument as described above.

Note that the Docker image already has a installation of Perple_X.

Additional analyses in R

Some analyses are more convenient to perform in R after the models have been run. For this purpose, the model saves a summary of key outputs to the _critical.csv file.

An R script for reading and working with this file is provided in the ./r directory.