[aspect-devel] [CIG-ALL] ASPECT 1.0 released

[Juliane Dannberg]

Hi,

I think it would be a really good idea to include such a self-consistent 
material model in ASPECT.
As Wolfgang already said, it is not a problem to have a material model 
based on a lookup table, in fact we already have such a material model 
in the current release (the Steinberger material model written by Rene 
Gassmoeller). This material model is used as an interface to data tables 
generated by the PERPLEX software. That means, technically it is no 
problem to read the material properties from a HeFESTo lookup table. One 
could follow the implementation in material_model/steinberger.cc and 
would need to consider things like the interpolation between different 
compositions (in case you have several compositional fields) and between 
data points of the input tables.

For the formulation of latent heat we included in ASPECT it should be no 
problem to use material properties from a lookup table. As far as I 
know, there are different approaches to include latent heat in the 
temperature equation. One would be to change the values of the thermal 
expansivity and the specific heat to effective values that also reflect 
latent heat effects, and this is how it is done in the Steinberger 
material model mentioned above. Another possibility is to add extra 
terms to the temperature equation that include the entropy changes, and 
this is what I implemented in ASPECT. So for this approach we need 
values for the entropy derivatives with respect to temperature and 
pressure, which could also be provided by a lookup table. When you 
implement a HeFESTo material model, you can decide which of these 
approaches you prefer to use.

If you would like to discuss this in more detail, Rene and myself (and I 
believe Timo as well) will also be at the CIDER meeting this summer. 
Maybe an interface between Burnman/HeFESTo and ASPECT could be a 
possible project there?

Best,
Juliane and Rene


On 04/23/2014 03:04 PM, Wolfgang Bangerth wrote:
>
> Hi Carolina,
> let me take this to the mailing list since we have more people there 
> with the relevant expertise.
>
>
>> I am a geodynamicist and have been involved in developing a 
>> self-consistent
>> thermodynamic framework for mantle composition and phases.
>> In the past myself and Lars Stixrude (the main author of HeFESTo, the
>> self-consistent thermodynamic code we use to predict phase assemblages
>> and physical properties of the mantle) we have talked ot CIG about
>> incorporating the thermodynamics and even releasing the code and 
>> lookup tables
>> through CIG. Mostly, time, moving to the UK and slowlgy replacing many
>> numerical recipes subroutine have not allowed us to fulfill that 
>> commitment.
>> In any case, given the way CitcomS was formulated it would not have 
>> been very
>> useful.
>>
>> However, this release of Aspect is very exciting and it seems to me 
>> could be
>> ready to explore the effects of thermodynamics on the flow. The mantle
>> has MANY phase transitions especially in the upper 800 km they are 
>> essentially
>> continuous. Across them there can be latent heat release or 
>> absorption but
>> also many changes in physical properties, both static and transport 
>> (thermal
>> conductivity and viscosity say, not something we can compute yet).
>> Some of are very sharp (VERY, i.e. hard for typical grids) happening 
>> across
>> less than 5 kms and some much more gradual. The effects on plate 
>> dynamics
>> can be quite important, including possibly intermittent layering, and 
>> also
>> strongly affect surface observables like dynamic topography.
>>
>> I’ve read the Aspect manual, admittedly quickly, and I wanted to ask 
>> you if
>> you think the release of Aspect is ready to interface with a 
>> thermodynamic lookup
>> table for given compositions. These tables include P, T, rho, 
>> Entropy, Cp,
>> Enthalpy, etc. Plus of course the shear and bulk moduli. It wasn’t 
>> clear to me
>> from
>> the manual whether it was yet possible for Aspect to include the
>> self-consistent thermodynamics or not.
>
> I think this should be possible, at the very least as far as the 
> material parameters are concerned. We parameterize material models in 
> the following way:
> - at every quadrature point, the core simulator evaluates pressure, 
> temperature and compositional fields
> - it then asks the material model in use for the corresponding 
> coefficients such as density, viscosity, thermal conductivity, etc
> - we have a few simply material models that compute these coefficients 
> as functions of p and T
> - but material models can also get the coefficients from lookup tables 
> and there is a model that Thomas Geenen has used that does exactly 
> this (I don't think we include it in the release because it's his 
> data, but you may ask him for the data or at least for the code -- 
> it's not very complicated).
>
> In other words, if all you want is to have a material model based on a 
> lookup table, then that's definitely possible and has been for a long 
> time. How complicated the lookup table is of course doesn't matter.
>
>
> Now, on to the second question: thermodynamics. My own understanding 
> of thermodynamics is very limited, so I may ask if Juliane could 
> possible chime in. Juliane implemented a model in which there is 
> release of latent heat, and you can see a simple benchmark on page 141 
> of the manual that illustrates this. I believe this video of hers
>   https://www.youtube.com/watch?v=dG-ULmcBr1E
> also uses this feature. In other words, I *think* what you need should 
> work, but I don't know enough about what parameterization you want to 
> use and how that fits into what is currently implemented. In any case, 
> however, we would certainly be very interested in seeing this made to 
> work, and we'd be happy to provide feedback and support if you or a 
> student wants to work on this!
>
> Best
>  Wolfgang
>
> where we simply ask the model to return to the core simulator
>