using dedalus for Rayleigh Benard convection

[Mishra, Shruti]

Hi All, 

I have a question about using Dedalus for Rayleigh Benard convection. I have come across a few references that use Dedalus for Ra in [1e4, 1e8] and Pr in [1e-1, 1e1]. Are there references / benchmarks for using Dedalus for Rayleigh Benard convection outside of this range? I am specifically interested in higher values of Ra. 

Thank you very much, 

Shruti 

[Daniel Lecoanet]

Hi Shruti,

This paper simulates Pr=1, Ra=1e9 in 3D: https://ui.adsabs.harvard.edu/abs/2018JFM...854R...3C/abstract

This paper simulates Pr=1, Ra=1e10 in 2D: https://ui.adsabs.harvard.edu/abs/2021MNRAS.508..132L/abstract

Higher Ra requires more computational resources. Roughly speaking, you have to increase the resolution by ~ a factor of 2 in each direction every time you increase Ra by a factor of 10.

Daniel

--
You received this message because you are subscribed to the Google Groups "Dedalus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dedalus-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/CAJywyQyuphv-dWNxHxO9tQJyrGixmZEYrhv0N1vrLvXK7ke2-g%40mail.gmail.com.

[Jeffrey S. Oishi]

Hi Shruti,

And the banner on the Dedalus website is 2D Rayleigh-Benard convection at Ra=1e11, which we made a movie of a few years back for the APS DFD Gallery of Fluid Motion:

https://www.youtube.com/watch?v=klntYmQtyE4

As Daniel implies but didn't directly say, the code is capable of reaching extreme Ra, P. We have tested the code for scaling up to at least O(1e4) cores without seeing any problems. It is a question of what computational resources you have available to you and how you want to use them. Keep in mind also that increasing the resolution by a factor of 2 in each dimension will also result in 2^3 increase in resolution for the spatial part and also an additional factor of 2 to satisfy the CFL condition, so a 2x increase in linear resolution will require 16x the computational resources. This is a limitation not of Dedalus but of numerical simulation generally. 

Jeff

Jeff

To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/F80453F8-0762-4906-B17D-30772F5DD47F%40northwestern.edu.

[Mishra, Shruti]

Hi Jeffrey and Daniel,

Thank you both very much. I will take a look at these references. May I ask what resolutions you used for 2D Rayleigh-Benard convection at Ra=1e11? 

Thank you,

Shruti

To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/CAAOZTN-_x39X%2BadtPkWd0Tz1j03O6E0ipX4ETKt8BRb2rWVYJQ%40mail.gmail.com.

[Jeffrey S. Oishi]

It's in the video.

To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-users/CAJywyQyQkjbZDoWcZrCDena9iJNNkpmckg8%3DCgDWorDzhNaYTQ%40mail.gmail.com.