Gravity Force

Lattice Boltzmann Method LBM: Theory
#1

Hi

In 3D15Q, I want to add the gravity term as a body force.
According to J.M.Buick and C.A. Greated, Gravity in a lattice Boltzmann model(PHYSICAL REVIEW vol61,5),
I have to add [D/(bc^2)Fe] to the distribution function.

F = [ rho(at each node) * g_lattice] or [ rho(at each direction of a node) * g_lattice] ??

Which is right? Or neither of which?

Thank you.

#2

It seems to me that none of the two formulas does an appropriate job. The inclusion of a body force in lattice Boltzmann has been discussed in-depth in a paper by Guo e.a. Through a Chapman-Enskog expansion, the paper shows that a simple force-term, linear to g_lattice, gives raise to unwanted error terms. To get rid of them, an additional force-velocity coupling must be implemented. The appropriate formula can be found in the paper.

#3

Hi Jonas,

In principal this is correct, but is it really necessary to do this for a constant and homogeneous body force?

Timm

#4

You are right, I am being too categorical with my answer: thank you for pointing this out. If the force field is space and time independent, YY’s force term is sufficient. Indeed, the error terms found in Guo’s Chapman-Enskog analysis are multiples of the time- resp. space-derivatives of the force. And… hmmm… now that I check the original post, I see that it is about gravity, which tends to be space and time independent.

So, the right way to implement the force term in this case is as follows. After collision, add the value F_i to each particle population:

F_i = rho t_i / c_s^2 c_i*g,

where rho is the density, t_i the lattice weight (e.g. 4/9, 1/9, 1/36 for D2Q9), c_s^2 the squared speed of sound (e.g. 1/3 for all standard lattices, including D3Q15), c_i the unitary lattice velocity ( e.g. (1,1) ) and g the force.

#5

Thank you for your reply.
In my case, gravity is constant. Certainly, according to Guo, Buick’s method is not accurate.
But it may be enough.
Anyway, you mean rho is density at node , not at each direction of the node.
Is it right ?

Thank you.

#6

That’s exactly right: rho(x,t) = sum_i f_i(x,t)

#7

I hope my observation may be of help to other people, since I am opening this topic again.

I was reading the Sukop book, at section 4.5 it reports how to implement the gravity force, as:

u_eq=u+tau*(rhog)/rho where obviously rhog is the gravity force but it may change according with the force in the system.

This definition is different from yours, but is it applicable for a 3D LBGK D3Q19 single phase problem?

Thanks in advance

Antonio

#8

Sukop’s implementation is not correct. I would stick to the articles by Guo or Khupershtokh (New method of incorporating a body force term into the lattice Boltzmann equation, Proc. 5th International EHD Workshop 2004).

#9

Thanks Timm for the answer.

I’ll read the journal and sort out how to implement this condition!

Kind Regards

Antonio