Lattice Boltzmann for Fluids

Presentation on theme: "Lattice Boltzmann for Fluids"— Presentation transcript:

1 Lattice Boltzmann for Fluids
Sauro Succi

2 LB For fluids The general idea of LB is to write down a set of
hyperbolic equations for a discrete set of movers (“propagons”) obeying a Propagation-Relaxation dynamics around prescribed local equilibria. Suitable choices of the equilibria lead to a broad variety of linear and non-linear PDE’s. A major advantage of the Prop-Relax dynamics is that it always proceeds along straight lines, no matter how complex the physics of the PDE’s. The method is especially useful for the Navier-Stokes fluid equations for quasi-incompressible fluids in d=1,2,3, which we now proceed to illustrate.

3 Navier-Stokes equations
Basic equations are known for nearly 2 centuries Reynolds number (L. Da Vinci)

4 Turbulence! h

5 Turbulent scales Kolmogorov length Faucet, Re=10^4, DOF=10^9
Car/Airpl, Re=10^6-8, DOF=10^14-18 Geo/Astro, Re=10^10, DOF=10^22

6 Energy spectrum: broad and gapless!

7 Navier-Stokes in d dimensions
Coordinate representation: Advection Pressure Dissipation # of Macrofields= Constraints:

8 Lattice Boltzmann: Platonic hydrodynamics
Triple infinity to just 19! Magic speeds! Exact sampling of frequent events

9 LB in d dimensions Lattice BGK (single-time relaxation) How to find:
c_i, Omega_ij , f_i^eq? Lattice BGK (single-time relaxation)

10 Navier-Stokes in d=2 dimensions
D2Q4=HPP (1976) NO! Natural: 4 speeds versus 6 macrofields: no way! D2Q6=FHP (1986) YES! Natural: 6 speeds versus 6 macrofields: ok!

11 Navier-Stokes in d=3 dimensions
6 speeds vs 10 fields: NO WAY! D3Q19: YES! 10 hydrofields + 9 “ghosts” YES!

12 Moment matching Moment matching proceeds exactly as in d=1, with
algebraic aggravation due to the tensorial structure. A crucial step is to recognize that the expansion is the lattice analogue of the Hermite expansion in continuum kinetic theory. This simplifies enormously moment-matching procedure and puts it on a systematic basis. Let us take a close lok at the procedure.

13 Moment matching: Mass Start from LB in differential BGK form:
Mass, sum over all discrete speeds: By imposing:

14 Moment matching: Momentum
Momentum: multiply by c_ia and sum over all discrete directions: By imposing: Is it Navier-Stokes? NOT YET! Constraints must be imposed on P_ab

15 Moment matching: momentum
The Navier-Stokes pressure tensor without Dissipation ( inviscid Euler regime) Advection + Pressure Dissipation is the macroscopic Manifestation of non-equilibrium!

16 Moment matching: Momflux
Momentum flux, multiply by c_{ia}c_{ib} and sum over i: Now because momflux is not a micro-invariant! How do we close this equation?

17 Boltzmann to Navier-Stokes
1. Scale separation: The molecular mean free path must be much smaller than any hydrodynamic length-scale (small Knudsen number) 2. Weak departure from local equilibrium: The Boltzmann probability distribution function must be close to a local Maxwell-Boltzmann T

18 Momflux: enslaving Momentum flux equation:
Enslaving: eliminate time derivative on a timescale tau=1/omega: Close to local equil: Dissipation Inertia+Pressure

19 Momflux: enslaving Injecting into the momentum equation, we obtain:
This must now match exactly the Navier-Stokes equations: Which implies the following tensorial constraints: Inertia+Pressure Dissipation

20 List of hydrodynamic constraints
Scalar: 1 constraint Vector: d constraints 2-Tensor: d*(d+1)/2 constraints 3-Tensor: d*(d+1)*(d+2)/6 constraints Q: Can we match the full list? How many discrete velocities?

21 Lattice equilibria: constraints
Continuity equation Momentum Equation Pressure+Advection Newtonian (linear) Dissipation

22 Lattice Equilibria Expand the local Maxwell-Boltzmann to second order in Mach number: where: Projection of the flow field over the i-th discrete speed is a lattice projector Question: How do we find the weights?

23 Lattice Equilibria By inserting
in the list of hydrodynamic constraints, we obtain: 0th order isotropy 2nd order isotropy 4th order isotropy

24 Lattice equilibria: Mass
Insert local equils: into Mass Conservation: Order 0: Normalization Order 1: Guaranteed by mirror symmetry (Parity invariance) Order 2: defines the lattice sound speed

25 Higher Order Lattice equilibria
Expand the local Maxwell-Boltzmann to third order in Mach number: Dipole Quadrupole Hexapole:

26 D1 lattice equilibria D1Q5

27 D2Q9 equilibria Quadratic polynomials (Check the algebra!):

28 Standard 3d LB lattices D3Q27=(D1Q3)^3 D3Q19= D3Q27-8 vertices

29 Higher order lattices D3Q39: order 6 isotropy D3Q93: order 8 isotropy

30 Very Higher Order Lattices
D=2, up to order 16! D=3, relativistic

31 Lattice vs continuum equilibria
Local Global Why not just take: Galilean invariance requires infinite series in the Mach number Infinite rank isotropic tensors, i.e. infinite connectivity!!! Scaling invariance v/vthermal: global to local very hard

32 Realizability (positivity): Low Mach

33 Summarizing: by choosing a suitable set of discrete
Summary Summarizing: by choosing a suitable set of discrete velocities and associated weights, ensuring fourth order isotropy, the LBE reproduces the Navier-Stokes equations for a fluid with: Eq. of state: Viscosity: The negative contribution (propagation viscosity) stems from second order expansion of the lattice streaming operator…

34 LBE assets + Streaming is linear and exact (no )
+ Non-linearity is local (round-off conservative) Laplacian-free dissipation + Pressure is local (weakly compressible) + Easy handling of complex geometries (straight lines) + Outstanding for parallel computing + Emergent complexity nearly for free + NOT limited to dilute gases !

35 LB across scales: from turbulence to biopolymers to quark-gluon plasma
Au

36 Boundary Conditions

37 No-slip flow via bounce-back
BUFFER WALL FLUID F F F

38 Bounce-Back B BUFFER WALL FLUID F F F

39 Assignements Write a D2Q9 code for channel flow and test the Poiseuille flow at different viscosities (see lb2.f) Same with a cylinder within the channel Same with a random porous media

40 End of the Lecture End of Lecture (for the detail thirsty,
see my book! 

41 Left over material

42 Equilibrium constraints
1 d d(d+1)/2 We need at least 1 + d + d*(d+1)/2 discrete velocities b> d*(1+(d+1)/2) = 5 movers in d=2 and 9 in d=3.

43 Dissipation-Viscosity
Expand in powers of Knudsen

44 Viscosity bounds Vanishing viscosity with dt=O(tau)=O(1) Stable up to

45 Lattice BGK equilibria
Expand the local Maxwell-Boltzmann (at least) to second order in Mach number: Dipole Quadrupole

46 Lattice BGK equilibria: Mom
Insert local equils: Deja’ vu: nothing new Deja’ vu Mirror symmetry

47 Lattice BGK equilibria: Momflux
Insert local equils: Mirror symmetry Deja’ vu Order 4: pressure and inertia

48 Lattice BGK equilibria: Dissipation
Insert local equils: Mirror symmetry Mirror symmetry Still order 4 isotropy: deja’ vu. Dissipation does not impose any high order constraint wrt pressure and inertia

49 Equation of State

50 LB Dissipation Usually: But LBGK has: hence

51 Boltzmann to NSE * Take kinetic moments: open hierarchy
* Closure: Weak departure from local equilibrium (Kn<<1) T u=u(r,t) T=T(r,t)

52 Dissipation This requires:

53 Discrete streaming Discrete streaming operator contains derivatives at all orders: For consistency with diffusion must be expanded to second order Propagation viscosity = -Delta t/2 < 0: CRUCIAL!

54 Viscosity Vanishing viscosity with dt=O(tau)=O(1) Stable up to
Beware: Large viscosities violate enslaving principle! Viscosity should be well below 1 (in LB units)