Concentration-dependent diffusivity in transport

[ecoon]

Currently ATS allows for "generalized concentration"-dependent diffusivity. It does this through an input spec that makes it sound like one is providing "primary species concentration" dependent diffusivity. And indeed, users are using it like they want to provide "diffusivity of O2(aq)" when this quantity is actually "diffusivity of the corresponding generalized flux". Reported by @scottpainter.

To fix this, we need to allow evaluators to have access to both primary and secondary species concentrations. Research is required to determine what exactly needs to be done.

1. in Amanzi's Alquimia PK -- how are "primary free ion concentration" and "secondary free ion concentration" dealt with? Are they always acquired from Alquimia? Only during diagnostics or every step?
2. Is there an evaluator (this would be a PrimaryEvaluator, manually managed by the Alquimia_PK) associated with these quantities?

These should be updated to be done every step, and saved via an evaluator.

At this point, a user could write a diffusivity that is a function of primary concentration or secondary concentration, but would be some effective summation of the true specie-dependent diffusivity.

Then an evaluator that (first derives and then) implements the primary-and-secondary specie-based sum to get a diffusivity for the generalized specie is needed.

[smolins]

ATS should only rely on concentrations passed in the Alquimia state for calculations, i.e. total concentrations. Primary and secondary (in aux output data) only used for output (https://github.com/LBL-EESA/alquimia-dev/blob/master/doc/api/APIv1_structures.rst#struct-alquimia-auxiliary-output-data). The use of the same diffusion coefficient for all species defining the component is common in reactive transport and usually not a problem. In fact the opposite is a problem. For example, see PFLOTRAN documentation https://documentation.pflotran.org/user_guide/cards/subsurface/chemistry_card.html namely, AQUEOUS_DIFFUSION_COEFFICIENTS List of primary species and diffusion coefficients for which the default aqueous diffusion coefficient will be overwritten. Species-dependent diffusion is not supported for simulations with aqueous complexation. (default units [m2/s]) The last sentence is because species dependent diffusion will not preserve electroneutrality. You need the multicomponent diffusion capabilities like CrunchFlow or PHREEQC. In CrunchFlow, it is only supported in the global implicit approach, not operator splitting, what we are using in ATS + Alquimia. What is the use case for this? One could try to set up the problem such that different components are defined for the species (now components) that need it, e.g. Sergi What is the use case?

…

On Tue, Jun 2, 2026 at 12:03 PM Ethan Coon ***@***.***> wrote: ecoon created an issue (amanzi/ats#345) Currently ATS allows for "generalized concentration"-dependent diffusivity. It does this through an input spec that makes it sound like one is providing "primary species concentration" dependent diffusivity. And indeed, users are using it like they want to provide "diffusivity of O2(aq)" when this quantity is actually "diffusivity of the corresponding generalized flux". Reported by @scottpainter. To fix this, we need to allow evaluators to have access to both primary and secondary species concentrations. Research is required to determine what exactly needs to be done. in Amanzi's Alquimia PK -- how are "primary free ion concentration" and "secondary free ion concentration" dealt with? Are they always acquired from Alquimia? Only during diagnostics or every step? Is there an evaluator (this would be a PrimaryEvaluator, manually managed by the Alquimia_PK) associated with these quantities? These should be updated to be done every step, and saved via an evaluator. At this point, a user could write a diffusivity that is a function of primary concentration or secondary concentration, but would be some effective summation of the true specie-dependent diffusivity. Then an evaluator that (first derives and then) implements the primary-and-secondary specie-based sum to get a diffusivity for the generalized specie is needed. — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS and Android. Download it today! You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[ecoon]

Thanks for this @smolins -- @scottpainter

The use case is faking gas diffusion in a vadose zone by first ensuring all aqueous gases are primary species and then increasing their diffusivity as a function of gas saturation.

[scottpainter]

The use case is vadose zone gas diffusion. We want to include that important transport mechanism to properly account for the dynamics of dissolved gases (which will partition to the gas phase and be subject to diffusion).
For species that partition to the gas phase, we can define an effective diffusion coefficient that combines the gas-phase and liquid-phase diffusion. However, this depends on henry's constant and so will be different for each component. Currently we assume equal diffusion coefficients, which allows us to solve the conservation equation for the generalized concentration without knowledge of the concentrations of the primary and secondary components. With unequal diffusion coefficients, we still solve the conservation equation for generalized concentration. However, the diffusive flux is no longer proportional to the gradient of the generalized concentration - it needs to be constructed from the gradients of the primary and secondary concentrations. There are no issues with this mathematically, but we currently do not have access to primary and secondary concentrations in the transport PK.

Note charge balance is not an issue. We're concerned with dissolved oxygen, carbon dioxide, and methane, which are all neutral species, so having them diffuse at different rates will not affect charge balance provided all the charged species diffuse at the same rate.

[scottpainter]

Leaving this issue open for now, but we are testing another simpler approach and will close this issue when that is finished.