Viscous interface: ldg-beta and ldg-tau definition

Cases
1

Hello,

where may I find more informations regarding ldg-beta and ldg-tau in the literature? Could you point to a good reference to study them?

Best regards

2

See Eq (25) and (27) in:

https://www.sciencedirect.com/science/article/pii/S0010465514002549

along with the references therein.

Regards, Freddie.

1 Like
3

Thanks a lot for the answer, very grateful

Best regards

4

Hi @fdw,

is there any relationship in terms of accuracy of the computed solution if the values are not 0.5 and 0.1 respectively? E.g. ldg-tau = 1,2 … or ldg-beta any other value in [-1, 1]?

Best regards

5

The differences are typically minor. Taking ldg-beta to be \pm 0.5 does have slight performance advantages hence it is the default.

Regards, Freddie.

6

To my understanding for ldc-beta ± 0.5 is because of upwind/downwind biasing with a compact stencil (am I right?) but what about tau? Is any value “reasonable” with minor deteriorations even large ones like 1.0, 2.0 etc…?

Best regards

7

The performance benefit comes from the fact that when beta is \pm 1/2 that you only need gradients from one side of the interface. This reduces the amount of work that the common interface flux code has to perform.

Larger values of ldg-tau may introduce excessive dissipation and there is not usually a benefit to using anything outside the range of zero to one.

Regards, Freddie.