[aspect-devel] Composition refinement strategy

[John Naliboff]

Hi Lev,

Sorry for the delay in replying and for the confusion regarding terminology. “Gradient” was probably the most confusing term I could have possibly used given that there is a mesh refinement criterion called “composition gradient” as Wolfgang noted.

My understanding of the “composition” refinement strategy is that it should refine in each compositional field near the boundaries of “different phases” (terminology taken from geomIO cookbook). If each of your compositional fields (crust, mantle, etc) is defined by having a value of 0 or 1, then it should refine within the field at the boundaries between values of 0 and 1 (i.e. where there is a gradient). 

I have not tried or looked in detail at the “composition_gradient” refinement function, but the idea is similar from the description in composition_gradient.cc. The difference lies in how the composition gradients are computed/defined. For those reading along, please do jump in If this not an accurate description.

Lev - I hope this helps clarify things and please do continue the discussion if further clarification is needed or you would like discuss the options further. In the short term, it may be worthwhile to see how the “composition” and “composition_gradient” options compare.

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis






> On May 17, 2017, at 5:51 PM, Lev Karatun <lev.karatun at gmail.com> wrote:
> 
> Hi John,
> 
> then I don't understand what the difference between "composition" and "composition gradient" strategies is. Can you clarify it please?
> 
> Best regards,
> Lev Karatun.
> 
> 2017-05-17 14:40 GMT-04:00 John Naliboff <jbnaliboff at ucdavis.edu <mailto:jbnaliboff at ucdavis.edu>>:
> Hi Lev,
> 
> The AMR should indeed be refining according to the compositional gradients. If you want to refine in regions without compositional gradients or another property (e.g. viscosity, velocity, etc), you can use the “minimum refinement function” option to initially refine based on position.
> 
> Cheers,
> John
> 
> *************************************************
> John Naliboff
> Assistant Project Scientist, CIG
> Earth & Planetary Sciences Dept., UC Davis
> 
> 
> 
> 
> 
> 
>> On May 17, 2017, at 12:20 PM, Lev Karatun <lev.karatun at gmail.com <mailto:lev.karatun at gmail.com>> wrote:
>> 
>> Hi,
>> 
>> I was using the composition refinement strategy for my models, and it's not working I expected it to. I have 5 c.f. in my model - upper and lower crust, weak zone, lithospheric and sub-lithospheric mantle. Lithoispheric mantle is shown on screenshot "composition_setup" (attached). The refinement scaling factors are as follows (in order): 
>> Compositional field scaling factors = 1,1,2,1,0
>> So upper, lower crust and the lithosphere have the same weight. However, when I look at the error indicator (and the resulting mesh), it looks like the refinement plugin prioritizes the boundaries of the c.f. over the interior (so basically it's looking at gradients, not c.f. value itself) -- see screenshot composition_error. 
>> 
>> Can someone please have a look and tell me if it's just me or if the plugin is indeed not working properly?
>> 
>> Thanks in advance!
>> 
>> Best regards,
>> Lev Karatun.
>> <composition_error.png><composition_setup.png>_______________________________________________
>> Aspect-devel mailing list
>> Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel>
> 
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel>
> 
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170518/8994eb60/attachment.html>