see the equations

if anything grad(k) should be the "limiter" in this case

I want to see the link where they justify this choice

Most likely is a typo. I dont find a reason why it should be that way, except when you have an explicit treatment of the divergence of one variable as a term in the equation of another variable

Yeah that's my impression too - I wonder why it forces you to specify the gradient field though, if in the overwhelming majority of cases you're just going to be feeding it the gradient of that field...

There are valid reasons to use different gradients

Exception I guess is if you're using linearUpwind limited

linearUpwind limited is as short way to say:

you can also say: linearUpwind corrected;

for the vectorized versions linearUpwind (linearUpwindV) you want to use grad(U) for example

finally, linearUpwind inherits from the NVD classes in FOAM

Oh true, forgot about those

which need the definition of the gradient

Yeah just seems like it wouldn't hurt for it to default to just grad(X) where X is the field you're specifying the scheme for

doesn't really matter in the end though

Extra specificity doesn't hurt I suppose when bytes are so cheap

Well you could avoid these confusions just by using the limited or corrected option

if you do:

linearUpwind grad(U);
...
snGrad(U) limited;

then the limiting is also applied right? Or is it not?

I assumed that it was

But never bothered to check

On the gradient, not on the divergence

On the divergence it will go to gradientSchemes and pick what you defined for grad(U) and then face interpolate

https://www.wolfdynamics.com/wiki/tipsandtricks.pdf Look at slide 21

That's not a justification. It could very well be a typo.

Yeah sure - just thought he wanted the link. Well in the beginning the mention "It is based on my personal experience". Otherwise i guess that people seem to agree that it is typo.

Hi I am modelling flow past a sphere in openFoam and would like to simulate the same in steady state...I tried using kw-sst-model but for a reynolds no of 400000 the drag coefficint is coming out to be 0.3 which does not match with the literature at all...which model should I use for such cases?

I am using steady state RANS for now

something like reynolds stress model or Spalart-Allmaras might be worth looking into