Geochemistry: reactive transport

[WilkAndy]

Reason

Implement reactive-transport in the geochemistry module to allow these kinds of simulations

Design

As i implement this functionality, i'll note the enhancements here since i'm currently not exactly sure what is required.

• [x] ModelGeochemicalDatabase struct needs to record the original basis, which will be used in the transport. EDIT: this is already in the original database reader, so is not needed
• [x] ModelGeochemicalDatabase needs to record the swap matrix to get back to the original basis. I think this should be a pointer to a DenseMatrix, rather than a ref, otherwise we'll be using a lot of memory at every node if there are no swaps involved. Probably can remove have_swapped_out_of_basis and have_swapped_into_basis. EDIT: no pointer, just a DenseMatrix() initially, and then resize appropriately when needed
• [x] introduce new flags basis_species_transported and eqm_species_transported and kin_species_transported into ModelGeochemicalDatabase that indicate whether the species are involved in transport. Off the top of my head, iff (a species is a mineral or surface_sorption_related=true) then species_transported = false.
• [x] In GeochemicalSystem add a getBulkTransportedMoles that returns the number of transported moles in the original basis
• [x] I think i can refactor GeochemicalSolver so that it accepts a GeochemicalSystem in the solveSystem method. The point of this is that a simulation can have just ONE GeochemicalSolver that does all the solves. I had initially thought, for some (good?) reason, that i'd need a GeochemicalSolver acting at each node, but that seems a waste of memory. Let's see if i can do the refactor....
• [x] Using the node void volume, translate the transported moles to a transported concentration for use by the kernels. I imagine this will be done inside a ReactiveTransport nodal UserObject.

Impact

Reactive transport in the geochemistry module!

[WilkAndy]

Hey @loganharbour , in a NodalUserObject _subproblem.mesh().getMesh().n_local_nodes() gives me the number of nodes that will be looped over in execute(). This is also the size of the std::vector<something> that i'm trying to compute with my NodalUserObject (ie, the "something" exists at each node).

• Presumably I will have to define a structure that maps _current_node->id() that is available in execute() to the correct index in my std::vector<something>

• What about coupledVars? Are they indexed by _current_node->id(), or something else. Viz, in execute(), should i be using var[_current_node->id()] or something else? EDIT: they have index [0].

• I might like to do the loop over all local nodes in initialSetup. If coupledVars are indexed by _current_node->id(), how do i know the index to use? EDIT: i'm pretty sure AuxVars values do not exist during the UserObject's initialSetup. This is a bit annoying - so far i'm sidestepping this issue by only allowing initially homogeneous distributions of reactants through the mesh (initial condition parameterised by Reals)

[WilkAndy]

Hi again @loganharbour . Can you give me some design advice?

At every node, i need a ModelGeochemicalDatabase. This holds the current basis species, the current equilibrium species, etc, that are relevant for that node. This is why i used a NodalUserObject in my "Reactor" classes. However, for reactive transport i see two possibilities, and i'm not really sure which is the best.

1. At every node, i also have a GeochemicalSystem. This holds all the information about the current molalities (concentrations) of every species. In the NodalUserObject, i update this (given information about sources/sinks active at the node, and using a GeochemicalSolver) at every time step. There may be AuxVariables that record some of these molalities - they just use the existing GeochemistyQuantityAux.

2. At every node I demand that there are AuxVariables that record the molalities (and any other info i need). Perhaps @YaqiWang 's VectorAuxVariable thing. I only have one GeochemicalSystem. In a loop over the nodes, i populate the single GeochemicalSystem using the AuxVariables at the current node, solve the system, and repopulate the AuxVariables with the solution.

The advantage of (1) is that it's pretty simple coding, since this was the design all along. It also allows users to be quite precise with their "constraints", eg fixing pH at one node but not another, although i'm not sure whether anyone will actually do this. However, it might be more complicated with mesh adaptivity, since i'm not relying on MOOSE doing the right thing with AuxVariables. Also, it's more difficult to restart a crashed simulation. Also, it uses more memory than (2).

If I go for (2), perhaps i shouldn't use a UserObject? I will need to const_cast the AuxVariables so i can directly write to them as soon as the GeochemicalSolver finds the solution. Perhaps i should use an AuxKernel instead that modifies a bunch of AuxVariables at once? But i still need a ModelGeochemicalDatabase at every node, which will be modified by the AuxKernel.

[WilkAndy]

I tried to do (2), above. It didn't work well because i couldn't actually write to the AuxVariables using the UserObject, boohoo. Not sure why - the const_cast was fine, but the AuxVariables didn't actually update.

(1) seems fine so far.

[WilkAndy]

I just completed my first reactive transport sim with geochemistry, @rpodgorney and @cpgr. Days of code-tidying to do, but the proof-of-concept is there - good way to end the week.

[rpodgorney]

Well done Andy. Enjoy your weekend.

-- Robert Podgorney, PhD, PG Geothermal Program Manager, Idaho National Laboratory Modeling and Simulation Lead, Utah FORGE Program Thermal Systems Co-Lead, INL Integrated Energy Systems Initiative

Office: 208-526-1524 Cell: 208-520-9361 email: robert.podgorney@inl.govmailto:robert.podgorney@inl.gov Statement of Confidentiality The contents of this e-mail message and any attachments are confidential and are intended solely for addressee. The information may also be legally privileged. This transmission is sent in trust, for the sole purpose of delivery to the intended recipient. If you have received this transmission in error, any use, reproduction or dissemination of this transmission is strictly prohibited. If you are not the intended recipient, please immediately notify the sender by reply e-mail or phone and delete this message and its attachments, if any.

From: Andy Wilkins notifications@github.com Reply-To: idaholab/moose reply@reply.github.com Date: Friday, August 7, 2020 at 1:23 AM To: idaholab/moose moose@noreply.github.com Cc: Robert Podgorney Robert.Podgorney@inl.gov, Mention mention@noreply.github.com Subject: [EXTERNAL] Re: [idaholab/moose] Geochemistry: reactive transport (#15693)

I just completed my first reactive transport sim with geochemistry, @rpodgorneyhttps://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Frpodgorney&data=02%7C01%7Crobert.podgorney%40inl.gov%7C9fcc0f4df09e4357baac08d83aa2cb62%7C4cf464b7869a42368da2a98566485554%7C0%7C0%7C637323818160662024&sdata=J8%2BoV6ldm0ED3PPrP%2BgQlUjnpFZonq1Rkac%2Bwk8RPvM%3D&reserved=0 and @cpgrhttps://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fcpgr&data=02%7C01%7Crobert.podgorney%40inl.gov%7C9fcc0f4df09e4357baac08d83aa2cb62%7C4cf464b7869a42368da2a98566485554%7C0%7C0%7C637323818160662024&sdata=q5EdQ4pLlxq%2B%2BOslVG40R2dprkraxz6GZwm%2BfCNbz0w%3D&reserved=0. Days of code-tidying to do, but the proof-of-concept is there - good way to end the week.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fidaholab%2Fmoose%2Fissues%2F15693%23issuecomment-670371774&data=02%7C01%7Crobert.podgorney%40inl.gov%7C9fcc0f4df09e4357baac08d83aa2cb62%7C4cf464b7869a42368da2a98566485554%7C0%7C0%7C637323818160672018&sdata=LTwjcPQF%2B1CIVr2ECuBjFfeHHj9rXEMptiYBcANTdzU%3D&reserved=0, or unsubscribehttps://gcc01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FACWGDT5E6EEIGFR7YZBLBV3R7OTXLANCNFSM4PSY4POQ&data=02%7C01%7Crobert.podgorney%40inl.gov%7C9fcc0f4df09e4357baac08d83aa2cb62%7C4cf464b7869a42368da2a98566485554%7C0%7C0%7C637323818160672018&sdata=ZIlc9TK2rju6gECamDsmrMMkZfAoVqEJJ0YpIzAXlT8%3D&reserved=0.