Runs for the manuscript

[EmiliaJarochowska]

I would like to propose the set of runs that we cover with all three implementation in the ms.
Let's focus on Euler scheme, 1e-6 dt, and the correct expression for the porosity diffusion coefficient, so phi ** 3, to be in agreement with the manuscript.
Here's a draft of planned runs for the paper. It includes

• a scan through the lengths of the grid size, with an increment of 125. These runs are now planned for 100 T* but might require longer runs until we see oscillations. Fortran runs of this have already been completed (1) by @csummers25 and are on Research Drive.
• a scan through values of porosity diffusion coefficient
• the usual benchmarks of (Phi0, Phi00) = {(0.6, 0.5), (0.8, 0.8)}

An editable version of this table is in Research Drive (a tab in Runs.odt), please can you look at it and see anything unnecessary or important missing? @csummers25 @cedrict
I am not sure yet how to adjust the number of points to the varying sizes of the grid. Always increase it by 50 nodes, so also by 25% of the initial size? E.g. 625 cm would be 250 nodes etc?

Phi0	Phi00	Run duration	b	Grid size	Rhytmite	Marlpde	Fortran
0.6	0.5	1T*	1.066667	500	0	0	0
0.8	0.8	100 T*	1.066667	500	0	0	0
0.8	0.8	100 T*	1.066667	625	0	0	0
0.8	0.8	100 T*	1.066667	750	0	0	0
0.8	0.8	100 T*	1.066667	875	0	0	0
0.8	0.8	100 T*	1.066667	1000	0	0	0
0.8	0.8	100 T*	1.066667	1125	0	0	0
0.8	0.8	100 T*	1.066667	1250	0	0	0
0.8	0.8	100 T*	1.066667	1375	0	0	0
0.8	0.8	100 T*	1.066667	1500	0	0	0
0.8	0.8	100 T*	1.066667	1625	0	0	0
0.8	0.8	30 T*	d_phi = 0.001	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.005	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.01	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.02	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.03	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.04	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.05	500	0	0	1
0.8	0.8	30 T*	d_phi = 0.1	500	0	0	1

We can do the runs of Fortran and marlpde (the latter requires some merging first), can I leave you rhytmite runs, @csummers25 once we agree on the parameters?

[csummers25]

For the changing grid size runs we want to keep the resolution the same, so we want to set nnx so that L/nnx stays constant. I can add a column to the table for the nnx values so that we make sure we use the same values?

I'm happy to take care of all the rhythmite runs, unless there's anything you think we should change I'll stick with the settings we've been using, which should match those on the repo. Just to be clear, there are two versions of the bottom boundary conditions that we've used, the scheme described in Discussion #5 and one where we override the value in the final zone so that it is the same as the nnx-1 zone, so that the spatial derivative is always zero. The first scheme gives oscillations that are more similar to the fortran ones, so that is what I've been using for the most part, is this also closer to what is used in marlpde?

[EmiliaJarochowska]

Marlpde uses "derivative": 0 which I reckon will end up the same thing as in rhytmite.
I linked the Oscillations_like_Fortran branch because this is where Hanno has been experimenting with bottom boundary conditions for solids (aragonite and calcite). He now has several variants tested but I think this commit would be the closest to rhytmite so I suggested merging this into main so we can use this as the final version for the runs.

Thanks for the offer of filling in the number of nodes :-) I think it is safer to have it written down indeed.

[EmiliaJarochowska]

For the implicit solver, Hanno requested that we use LSODA.

[EmiliaJarochowska]

I saw that runs with increased resolution (L = 500 and nnx = 300, 400) had been made with a dt of $10^{-7}$. Shall we re-do it with dt $10^{-6}$ to match the time step in other runs and examine one aspect at a time?
I would also like to remove the Fortran runs from the resolution folder and re-run these because of the most recent correction with removing multiplication by $\sqrt(K_c)$. Just a heads up in case anyone wonders. I will wait with removing them until tomorrow.