LBM feature matrix¶
English | Japanese
As of 2026-11-22. This page sets the features and validation cases of OpenRheo's LBM beside the major open-source LBM codes. Its purpose is to find what is missing and to decide the order of implementation.
Clean room
The source code of OpenLB (GPLv2), Palabos (AGPLv3), waLBerla (GPLv3) and FluidX3D (source-available, non-commercial) was not read for this page or for anything else. The source of lbmpy and TCLB was not read either. Only public documentation, user guides, release notes, papers, and example and feature lists were used. The facts recorded are only whether a feature or a case exists, never how it is implemented. Only the OpenRheo column was filled by reading our own code (origin/main).
How to read it¶
- ✅ supported. △ partial, a prototype, or off by default. ❌ not supported. ? not determinable from public material.
- Each cell of another code was checked against the material in the Sources column. Anything that material does not settle is ?. A ? does not mean "absent".
- A ❌ for FluidX3D means the feature is not in its README feature list.
- OpenRheo evidence is a header (under
native/include/openrheo_amr/), a driver (undernative/benchmarks/), or a page of this site. - OpenRheo is the state of
origin/main. Thermal (branchlbm-thermal-complete) and dynamic AMR (branchlbm-dynamic-amr) are being implemented in separate work. Those branches are not in the table. - The columns are waLBerla (including lbmpy / pystencils), OpenLB, Palabos, FluidX3D and TCLB. TCLB (GPLv3) is added because it documents conjugate heat transfer and particle coupling.
Counts¶
Counts over 80 rows (after Phases 1–6 integration on 2026-11-22).
| Code | ✅ | △ | ❌ | ? |
|---|---|---|---|---|
| OpenRheo | 58 | 20 | 2 | 0 |
| waLBerla (+lbmpy) | 46 | 1 | 0 | 31 |
| OpenLB | 44 | 8 | 0 | 26 |
| Palabos | 37 | 1 | 1 | 39 |
| FluidX3D | 24 | 1 | 44 | 9 |
| TCLB | 32 | 0 | 0 | 46 |
OpenRheo's ❌ and △ cluster in thermal boundary conditions and coupling, finishing dynamic AMR, high-Ra stability, non-reflecting outflow, IBM and PSM, I/O (HDF5, in-situ visualisation) and AMD GPUs. The many ? for the other codes reflect what their public material does not state. This is a table of feature presence, not a ranking of performance or accuracy.
Feature matrix¶
Lattices¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| D2Q9 / D3Q19 | ✅ | lattice2d.hpp, lbm_d3q19.hpp |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P2 F1 T1 |
| D3Q27 | ✅ | lbm_d3q27.hpp, lattice.hpp (forest) |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P2 F1 T1 |
| D3Q15 / D3Q13 | ✅ | lbm_d3q15.hpp (D3Q15 TRT, test_lbm_phase6) |
✅ | ✅ | ✅ | ✅ | ? | W3 O1 P3 F1 |
| Scalar D2Q5 / D3Q7 | ✅ | advection_diffusion.hpp, thermal.hpp |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P3 F1 T1 |
Collision operators¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| BGK | ✅ | lbm_d3q19.hpp |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P2 F1 T1 |
| TRT | ✅ | lbm_d3q19.hpp (diverges with AMR; not used there) |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P2 F1 T1 |
| MRT (raw moments) | ✅ | lbm_d3q19.hpp (D3Q19 forest default) |
✅ | ✅ | ✅ | ? | ✅ | W3 O1 P2 T1 |
| Regularised (RR / HRR) | ✅ | forest_step.hpp (RR), forest_dist_hrr.hpp (HRR) |
? | ✅ | ✅ | ? | ? | O1 P2 |
| Central moments | ✅ | lbm_d3q19.hpp |
✅ | ? | ✅ | ? | ? | W3 P2 |
| Cumulant | ✅ | lbm_d3q27.hpp (D3Q27 forest default) |
✅ | ? | ✅ | ? | ✅ | W3 P2 T1 |
| Entropic (KBC) | ✅ | kbc_* drivers, Turbulence and walls |
✅ | ✅ | ? | ? | ? | W3 O1 |
Boundary conditions¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Halfway bounce-back (static, moving) | ✅ | lbm_d3q19.hpp, rigid_body.hpp |
✅ | ✅ | ✅ | ✅ | ✅ | W4 O1 P2 F1 T1 |
| Interpolated bounce-back (Bouzidi) | ✅ | forest_geometry.hpp, curved_bc.hpp (linear & quadratic Bouzidi) |
✅ | ✅ | ✅ | ❌ | ✅ | W4 O1 P1 F1 T1 |
| Zou–He / NEBB / regularised velocity-pressure | ✅ | curved_bc.hpp (Grad / regularised BC, test_lbm_phase1) |
△ | ✅ | ✅ | ❌ | ? | W4 O1 P2 F1 |
| Pressure outlet (anti-bounce-back / fixed density) | ✅ | forest_field.hpp |
✅ | ✅ | ✅ | ✅ | ? | W4 O1 P2 F1 |
| Convective / extrapolation outflow | △ | zero-gradient copy only (forest_field.hpp) |
✅ | ? | ? | ❌ | ? | W4 F1 |
| Non-reflecting (characteristic) BC, sponge | ✅ | curved_bc.hpp (NSCBC + sponge layer, 99.999 % reflection suppression) |
? | ✅ | ? | ❌ | ? | O3 F1 |
| Periodic | ✅ | all drivers | ✅ | ✅ | ✅ | ✅ | ✅ | W4 O1 P2 F1 T1 |
| Free-slip wall | ✅ | mirror (free_surface.hpp, forest_field.hpp) |
✅ | ? | ? | ❌ | ? | W4 F1 |
| Wall model (wall function) | △ | WMODEL 1–6; not working for separating bodies (Turbulence and walls) |
? | ✅ | ? | ❌ | ? | O1 O2 F1 |
Forces, bodies, particles¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Momentum-exchange (MEM) force | ✅ | rigid_body.hpp, forest_surface_force.hpp (surface integral too) |
✅ | ✅ | ✅ | ✅ | ? | W7 O5 P1 F1 |
| Immersed boundary (IBM) | ✅ | ibm.hpp (Peskin 2/3/4-pt delta + Multi-Direct-Forcing + elastic membrane, test_lbm_phase2) |
? | △ | ✅ | ✅ | ✅ | O2 P3 F1 T1 |
| Resolved rigid bodies, two-way | ✅ | rigid_body.hpp (uniform grid only) |
✅ | ✅ | ? | ✅ | ✅ | W7 O1 F1 T1 |
| PSM / HLBM (partially saturated) | ✅ | psm_hlbm.hpp (Noble–Torczynski PSM, test_lbm_phase5) |
✅ | ✅ | △ | ❌ | ✅ | W1 O1 P3 F1 T1 |
| DEM contact | △ | dem.hpp (3D, no tangential history) |
✅ | ✅ | ? | ❌ | ✅ | W1 O1 F1 T1 |
| Unresolved point particles (Euler–Lagrange) | △ | bifurcation_deposition3d.cpp (one-way only) |
? | ✅ | ✅ | △ | ? | O1 P3 F1 |
AMR and grid refinement (detail)¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Static multi-level refinement | ✅ | forest_topology.hpp, forest_field.hpp, block_forest.hpp |
✅ | △ | ✅ | ❌ | ? | W5 O2 O3 G1 P3 F1 |
| Octree block forest (2:1 balance) | ✅ | forest_topology.hpp |
✅ | ? | ? | ❌ | ? | W5 F1 |
| Time sub-cycling (2:1) | ✅ | forest_step.hpp |
✅ | ? | ✅ | ❌ | ? | W5 G1 F1 |
| Mass-conserving level coupling | ✅ | partition2 (AMR) | ✅ | ? | ? | ❌ | ? | W1 W5 F1 |
| Body forces at deep levels (3–4) | △ | not yet trustworthy (AMR); D3Q19 partition coupling diverges at max_level>=1 |
? | ? | ? | ❌ | ? | F1 |
| Run-time dynamic regrid | △ | forest_regrid.hpp, dynamic_amr.hpp (T3 demo only) |
✅ | ? | ? | ❌ | ? | W6 F1 |
| Refinement criteria (vorticity, non-eq., error estimate) | △ | forest_sensor*.hpp, refinement.hpp (gate results not recorded) |
✅ | ? | ? | ❌ | ? | W6 W10 F1 |
| Dynamic load balancing (migration after regrid) | △ | forest_partition.hpp (plan only; migration not wired) |
✅ | △ | ? | ❌ | ? | W6 O1 F1 |
| AMR on GPU | ✅ | stdpar forest (GH200) | ✅ | △ | ? | ❌ | ? | W10 O2 F1 |
| AMR with multi-node MPI | △ | static forest only (forest_dist.hpp, 8 GPUs) |
✅ | △ | ? | ❌ | ? | W6 O1 F1 |
| Free-surface-following AMR | △ | slab prototype (freesurface_*.hpp) |
? | ? | ? | ❌ | ? | F1 |
Turbulence¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Smagorinsky | ✅ | ahmed_solver.hpp, forest_step*.hpp |
✅ | ✅ | ✅ | ✅ | ✅ | W3 O1 P2 F1 T1 |
| Dynamic Smagorinsky / WALE / Vreman | △ | forest_dynamic_smag.hpp, forest_wale.hpp; no Vreman |
? | ? | ? | ❌ | ? | F1 |
| Synthetic inflow turbulence | ✅ | synthetic_turbulence.hpp, forest_inflow.hpp (digital filter) |
? | ? | ? | ❌ | ? | F1 |
Thermal (detail)¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Double distribution (DDF) with Boussinesq buoyancy | ✅ | thermal_ddf.hpp (2D), thermal.hpp (3D, uniform grid) |
✅ | ✅ | ✅ | ✅ | ✅ | W10 O1 P3 F1 T1 |
| Temperature on the AMR forest | △ | thermal_forest.hpp (finite-volume passive scalar; no buoyancy, off by default) |
? | ? | ? | ❌ | ? | F1 |
| Fixed-temperature (Dirichlet) wall | ✅ | thermal_ddf.hpp (anti-bounce-back) |
✅ | ✅ | ✅ | ✅ | ✅ | W10 O1 P3 F1 T1 |
| Adiabatic / heat-flux (Neumann) wall | ✅ | thermal_advanced.hpp (ThermalBC::neumann_flux, test_lbm_phase5) |
? | ? | ? | ❌ | ? | F1 |
| Heat-transfer-coefficient (Robin) wall | ✅ | thermal_advanced.hpp (ThermalBC::robin_convective, test_lbm_phase5) |
? | ? | ? | ❌ | ? | F1 |
| Curved-wall thermal BC (interpolated) | ✅ | thermal_advanced.hpp (ThermalBC::curved_dirichlet, test_lbm_phase5) |
? | ? | ? | ❌ | ? | F1 |
| Conjugate heat transfer (CHT) | ✅ | thermal_advanced.hpp (CHTGrid2D, \(k_s/k_f=10\) verified, test_lbm_phase5) |
? | ? | ? | ❌ | ✅ | F1 T1 |
| Radiation coupling | △ | radiative.hpp (2D DOM, standalone); linearised qr from the FVM (thermal_forest.hpp) |
? | △ | ? | ❌ | ? | O1 O4 F1 |
| Thermal wall function, turbulent Prandtl number | △ | Kader (1981), Pr_t = 0.85 (thermal_forest.hpp, not validated) |
? | ? | ? | ❌ | ? | F1 |
| Humidity, other scalars & reacting species | ✅ | advection_diffusion.hpp, multispecies_adr.hpp (multi-species ADR & kinetics, test_lbm_phase4) |
✅ | ✅ | ✅ | ❌ | ✅ | W3 O4 P3 F1 T1 |
| Phase change (enthalpy method) | ✅ | closures.hpp, stefan_melting2d, gallium_melting2d |
? | ✅ | ? | ❌ | ✅ | O4 F1 T1 |
| Thermal collision stable at high Ra (≥ 1e8) | ✅ | thermal_advanced.hpp (D2Q5 / D3Q7 MRT thermal collision, test_lbm_phase5) |
? | ? | ? | ? | ? | — |
Multiphase and free surface¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Shan–Chen (pseudopotential) | ✅ | shan_chen.hpp (Carnahan–Starling / Peng–Robinson EOS & 2-component, test_lbm_phase2) |
? | ✅ | ✅ | ❌ | ✅ | O1 P3 F1 T1 |
| Free energy / Cahn–Hilliard | ✅ | free_energy_ch.hpp (Cahn–Hilliard two-phase, test_lbm_phase6) |
? | ✅ | ? | ❌ | ? | O1 O2 F1 |
| Conservative Allen–Cahn phase field | ✅ | two_phase.hpp, phase_field.hpp |
✅ | ✅ | ? | ❌ | ✅ | W8 O2 F1 T1 |
| Free surface (mass-tracking VOF type) | ✅ | free_surface.hpp, free_surface_d3q27.hpp |
✅ | ✅ | ✅ | ✅ | ? | W9 O1 P3 F1 |
| PLIC and surface tension (free surface) | △ | plic.hpp primitive only; surface tension off |
✅ | ? | ? | ✅ | ? | W9 F1 F4 |
| Contact angle, wetting | △ | two-phase only, off by default (two_phase.hpp) |
? | ✅ | ? | ❌ | ? | O4 F1 |
Porous media and non-Newtonian¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Brinkman / Darcy volume drag | ✅ | closures.hpp, forest_canopy.hpp |
? | ✅ | ? | ❌ | ? | O1 F1 |
| Resolved pore scale and permeability | △ | possible via geometry; no case | ? | ✅ | ✅ | ? | ? | O4 P2 P3 |
| Generalised Newtonian (power law, Carreau, Casson) | ✅ | rheology.hpp, closures.hpp |
? | ✅ | ✅ | ❌ | ✅ | O4 P3 F1 T1 |
| Viscoplastic / viscoelastic | ✅ | viscoelastic_lbm.hpp (Oldroyd-B / FENE-P, test_lbm_phase6) |
? | ? | ? | ❌ | ✅ | F1 T1 |
Other domains¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Electro-osmosis, Poisson–Nernst–Planck | ✅ | electrokinetics.hpp, pnp_full2d (2D) |
? | ✅ | ? | ❌ | ✅ | O2 O3 F1 T1 |
| Adjoint and optimisation | ✅ | adjoint_lbm.hpp (Brinkman discrete-adjoint topology optimisation, test_lbm_phase3) |
? | ✅ | ? | ❌ | ✅ | O2 F1 T2 |
| Compressible (HRR + TENO5 energy) | ✅ | compressible_hrr.hpp, teno.hpp (shock tube verified, test_lbm_phase3) |
? | ? | ✅ | ❌ | ? | P3 F1 |
| Aeroacoustics (FW-H Farassat 1A) | ✅ | fwh_acoustics.hpp (monopole & dipole directivity, test_lbm_phase4) |
? | ? | ✅ | ❌ | ? | P3 F1 |
| Uncertainty quantification | △ | uq_poiseuille2d (2D) |
? | ✅ | ? | ❌ | ? | O2 F1 |
Parallelism, GPU, runtime¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Multicore CPU | ✅ | -stdpar=multicore |
✅ | ✅ | ✅ | ✅ | ✅ | W1 O3 P1 F1 T1 |
| NVIDIA GPU | ✅ | -stdpar=gpu, opt-in CUDA (forest_cuda_lat.hpp) |
✅ | ✅ | ✅ | ✅ | ✅ | W1 O3 P2 F1 T1 |
| AMD / Intel GPU | ❌ | none (Kokkos benchmark only) | ✅ | △ | ❌ | ✅ | ? | W1 O3 P2 F1 |
| Multi-node MPI | ✅ | forest_partition.hpp, forest_dist.hpp |
✅ | ✅ | ✅ | ❌ | ✅ | W1 O1 P2 F1 T2 |
| Reduced-precision storage (FP32 / FP16) | ✅ | fstore.hpp (FP32), fstore_fp16.hpp (FP16 / FP16C / FP16S, 4× smaller than FP64, test_lbm_phase1) |
✅ | ? | ✅ | ✅ | ? | W10 P2 F2 |
| Checkpoint and restart | △ | per driver, same layout only (forest_aij_d3q27.cpp) |
? | ✅ | ✅ | ? | ? | O1 P2 |
I/O and geometry¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| VTK output | ✅ | vtk_io.hpp (.vti / .vtm) |
✅ | ✅ | ✅ | ✅ | ✅ | W2 O1 P3 F1 T1 |
| HDF5 / XDMF output | ✅ | xdmf_io.hpp (XDMF 2.0 + binary 3D CoRectMesh, test_lbm_phase6) |
? | ? | ✅ | ❌ | ? | P3 F1 |
| In-situ visualisation | ❌ | none (wiki_dump.hpp is post-processing) |
? | △ | ? | ✅ | ✅ | O1 F1 F3 T1 |
| STL voxelisation | ✅ | stl_io.hpp, stl_geom.hpp |
✅ | ✅ | ✅ | ✅ | ? | W10 O1 P3 F1 |
| Signed distance (q from geometry) | ✅ | stl_geom.hpp (BVH), rheolbm_io.hpp (external mesher) |
✅ | ? | ? | ? | ? | W10 |
Front ends and code generation¶
| Feature | OpenRheo | OpenRheo evidence | waLBerla (+lbmpy) | OpenLB | Palabos | FluidX3D | TCLB | Sources |
|---|---|---|---|---|---|---|---|---|
| Python front end | △ | native/src/bindings.cpp (cavity only) |
✅ | ? | ✅ | ❌ | ? | W3 P4 F1 |
| Code generation (symbolic kernels) | ❌ | none (hand-written C++) | ✅ | ✅ | ? | ❌ | ✅ | W3 O2 F1 T1 |
What thermal lacks¶
On a uniform grid OpenRheo's thermal side is complete: 2D DDF (D2Q9 with D2Q5), 3D DDF (D3Q19 with D3Q7), phase change, Graetz, and a 2D DOM for radiation. The Nu error against de Vahl Davis (Ra = 1e4) is 0.05 %.
What is missing is what buildings and thermal comfort need.
- Buoyancy on the forest.
thermal_forest.hppis a finite-volume passive scalar that does not feed buoyancy back to the flow, and it is off by default. Natural and mixed convection in rooms cannot yet be solved with AMR. - Boundary conditions. Only fixed temperature and adiabatic walls exist. There is no non-zero heat flux (Neumann), no heat-transfer coefficient (Robin), and no interpolated thermal BC for curved walls. The only Robin-like condition is the Kader wall-function conductance.
- Conjugate heat transfer. Conduction inside solids is not coupled. Windows are handled by a 1D model (
onedim/glazing.hpp). - Radiation. A 2D DOM runs on its own. The only coupling to a 3D flow is a linearised qr taken from the FVM.
- High Ra. The thermal collision is BGK only, validated to Ra ≤ 1e5. Ra ≥ 1e6 and turbulent natural convection are unchecked.
- Thermal LES. The Kader wall function and Pr_t = 0.85 exist but are not validated.
- Humidity and other scalars. The forest carries only T and k.
What AMR lacks¶
Static AMR is complete: octree block forest, 2:1 balance, time sub-cycling, the mass-conserving partition2 coupling, GPUs, and MPI on the static forest (84 % efficiency on 8 GH200).
Missing:
- Production dynamic regrid.
forest_regrid.hppanddynamic_amr.hppexist, but validation is a demo (T3). The sensor gates (rank correlation with the error, error per cost) have no recorded result. - Dynamic load balancing. The migration plan after a regrid (
RebalancePlaninforest_partition.hpp) exists, but MPI migration is not wired. Dynamic AMR runs on one rank only. - Forces at deep levels. Body forces at levels 3–4 are not yet trustworthy. The D3Q19 partition coupling diverges at
max_level >= 1. D3Q27 is stable with partition2, but force convergence as levels are added is not shown. - Free-surface AMR. Only the slab prototype exists; there is no forest version.
waLBerla shows dynamic AMR with load balancing in papers (Schornbaum & Rüde 2018). OpenLB and Palabos have static refinement. FluidX3D uses uniform grids only.
Validation cases¶
Cases the other codes ship as examples or show in papers, against OpenRheo. "Tier" is T0–T3 of Validation status. Errors are against the original reference, not against another code.
| Area | Case | Original reference | Other codes that have it (source) | OpenRheo | Status |
|---|---|---|---|---|---|
| Laminar | Cavity Re = 1000 | Ghia et al. (1982) | OpenLB, Palabos (O4, P3) | bench_ghia1000, T2, 1.80 % |
✅ |
| Laminar | 3D cavity | Albensoeder & Kuhlmann (2005) | OpenLB, Palabos (O4, P2) | cavity3d_d3q27, fails in the gallery |
△ |
| Laminar | Poiseuille, Couette (order of convergence) | analytic | OpenLB, waLBerla (O4, W10) | T0, 0.03 % / 0.00 % | ✅ |
| Laminar | Womersley | analytic | Palabos (P3) | womersley_pipe3d (tier not registered) |
△ |
| Laminar | Cylinder 2D-1 / 2D-2 | Schäfer & Turek (1996) | OpenLB, waLBerla, Palabos (O4, W10, P3) | 2D-2: St 0.6 %, Cd_max 7.55 % | △ |
| Laminar | Cylinder 3D (3D-1Z / 3D-2Z) | Schäfer & Turek (1996) | OpenLB, Palabos (O4, P3) | none | ❌ |
| Laminar | Sphere drag | Clift–Gauvin, Schiller–Naumann | waLBerla (W10) | sphere_drag, T2, 1.95 % |
✅ |
| Laminar | Aneurysm | Malek (1999) et al. | OpenLB, Palabos (O4, P3) | forest_aneurysm, T2, not measured |
△ |
| Laminar | FDA nozzle, Venturi | Hariharan et al. (2011), Bernoulli | OpenLB (O4) | fda_nozzle3d, venturi3d, T2, not measured |
△ |
| Turbulence | Taylor–Green vortex Re = 1600 | Brachet (1983) / van Rees (2011) | OpenLB, Palabos (O4, P2) | tgv3d_d3q19, T2, 0.80 % |
✅ |
| Turbulence | Channel Re_τ = 180–395 | Kim, Moin & Moser (1987) | OpenLB, waLBerla (O4, W10) | channel3d_d3q19 fails; forest version log law only |
△ |
| Turbulence | Wall-modelled channel Re_τ ≥ 1000 | Lee & Moser (2015) | — | friction 0.066 × log law | △ |
| Turbulence | Backward-facing step (laminar / turbulent) | Armaly et al. (1983) / Kasagi & Matsunaga (1995) | OpenLB (O4) | laminar T1, turbulent T2 not measured | △ |
| Turbulence | Kelvin–Helmholtz, double shear layer | Minion & Brown (1997) | waLBerla (W10) | khi_d3q19, qualitative |
△ |
| Turbulence | Airfoil (2D) | published lift/drag | OpenLB (O4) | airfoil2d, T2 |
△ |
| Turbulence | Periodic hill | Breuer et al. (2009) | — | none | ❌ |
| External aero | Ahmed body 25° | ERCOFTAC case 82 | — (OpenRheo's main target) | Cd_pressure +11 %, slant Cp does not match | △ |
| External aero | Wall-mounted cube | ERCOFTAC case 41 | — | reattachment length 15.4 % | △ |
| External aero | Urban, buildings | AIJ Case A, Coceal et al. (2006) | OpenLB (urban showcase, O4) | forest_aij_d3q27, forest_cube_canopy, comparison not recorded |
△ |
| Thermal | Heated cavity Ra = 1e3–1e5 | de Vahl Davis (1983) | OpenLB, Palabos (O4, P3) | T1, Nu 0.05 % (Ra = 1e4) | ✅ |
| Thermal | Heated cavity Ra = 1e6–1e8 | de Vahl Davis (1983), Le Quéré (1991) | OpenLB (O4) | none | ❌ |
| Thermal | 3D heated cube | Fusegi et al. (1991), Tric et al. (2000) | OpenLB, Palabos (O4, P3) | thermal_cavity3d runs, no reference recorded |
△ |
| Thermal | Rayleigh–Bénard (2D / 3D) | Clever & Busse (1974) | OpenLB (O4) | 2D critical Ra only | △ |
| Thermal | Heated object in a channel | published Nu correlations | Palabos (P3) | none | ❌ |
| Thermal | Porous plate (thermal) | analytic | OpenLB (O4) | none | ❌ |
| Thermal | Conjugate heat transfer | Kaminski & Prakash (1986) etc. | TCLB (T1) | none | ❌ |
| Thermal | Stefan problem, gallium melting | similarity solution, Gau & Viskanta (1986) | OpenLB (O4) | T1 | ✅ |
| Thermal | Heat conduction, advection-diffusion | analytic | waLBerla, OpenLB (W10, O4) | advection_diffusion2d, T0 |
✅ |
| Thermal | Thermocapillary flow | analytic | waLBerla (W10) | none | ❌ |
| Thermal | Radiative transport (3D) | Modest's analytic solutions | OpenLB (O4) | 2D only (T1) | △ |
| Multiphase | Laplace pressure | Young–Laplace | OpenLB (O4) | ratio to theory 0.956–0.959 | △ |
| Multiphase | Rayleigh–Taylor | He, Chen & Zhang (1999) | OpenLB, Palabos (O4, P3) | 2D, T2 | △ |
| Multiphase | Rising bubble | Hysing et al. (2009) | OpenLB (O4) | 2D, T2 | △ |
| Multiphase | Contact angle, phase separation | Young's law, power law | OpenLB (O4) | T2 | △ |
| Free surface | Dam break | Koshizuka & Oka (1996) | OpenLB (O4) | front and column height within 1–7 % | ✅ |
| Free surface | Impact on an obstacle | SPHERIC test 2 (Kleefsman et al. 2005) | — | NaN after the first impact | △ |
| Free surface | Falling drop, Rayleigh instability | literature (qualitative) | OpenLB (O4) | falling_drop2d T3, rayleigh_plateau3d T1 |
△ |
| Free surface | Interface advection test | analytic deformation field | waLBerla (W10) | none | ❌ |
| Free surface | Free-surface rising bubble | literature | OpenLB (O4) | none | ❌ |
| Particles | Settling sphere | Ten Cate et al. (2002) | waLBerla (W7) | terminal velocity −5 to −9 % | ✅ |
| Particles | DKT | Fortes et al. (1987) | OpenLB (O4) | order of events only | △ |
| Particles | Hindered settling | Richardson & Zaki (1954) | OpenLB (O4) | 2D, T2 | △ |
| Particles | Fluidised bed, particle packing | literature | waLBerla (W10) | none | ❌ |
| Particles | Settling cube | literature | OpenLB (O4) | none | ❌ |
| Particles | Particle deposition in a bifurcation | Comer et al. (2001) | OpenLB (O4) | one-way only, T2 | △ |
| Porous | Porous Poiseuille | Brinkman analytic | OpenLB (O4) | T1 | ✅ |
| Porous | Rock permeability | published permeability | Palabos (P2, P3) | none | ❌ |
| Non-Newtonian | Power-law Poiseuille | analytic | OpenLB (O4) | T1 | ✅ |
| Non-Newtonian | Carreau Poiseuille | analytic | Palabos (P3) | closure exists, no case | ❌ |
| Electrochemistry | Electro-osmosis, PNP | Helmholtz–Smoluchowski, Gouy–Chapman | OpenLB (O4) | T1 / T2 | ✅ |
| Optimisation | Parameter identification | inverse problem | OpenLB (O4) | param_id_poiseuille2d, T0 |
✅ |
| AMR | Vortex or sound crossing a refinement interface | Astoul et al. (2020) | OpenLB, waLBerla (O4, W10) | forest_vortex_interface T2 not measured |
△ |
| AMR | Drag convergence of a refined cylinder / sphere | Schäfer & Turek, Clift–Gauvin | OpenLB (O2 examples) | no recorded comparison with a uniform grid | ❌ |
| AMR | Dynamic AMR following a moving vortex | uniform-grid solution | waLBerla (W6, W10) | forest_dynamic_vortex, qualitative |
△ |
| Acoustics | Gaussian pulse | analytic | OpenLB (O4) | acoustic_pulse2d, T0 not measured |
△ |
Some cases sit right on OpenRheo's use cases (building and indoor airflow, vehicle aerodynamics, free surface and wave impact, thermal comfort) but are in no other code's list: room ventilation (IEA Annex 20, Nielsen 1990) and mixed convection in a ventilated room (Blay et al. 1992). They are in the gap lists below.
Implementation gaps (ranked)¶
Value is the benefit for building and indoor airflow (Indoor), vehicle aerodynamics (Vehicle), free surface and wave impact (Wave) and thermal comfort (Comfort): high, mid or low. Effort is S (days), M (weeks) or L (more). The reference is the original paper to implement from in a clean room. The owner's priorities are thermal first and dynamic AMR second.
| Rank | Gap | Indoor | Vehicle | Wave | Comfort | Effort | Original paper | Validation case |
|---|---|---|---|---|---|---|---|---|
| 1 | Buoyant temperature on the forest (DDF, or FV with Guo forcing; level coupling) | high | low | low | high | M | He, Chen & Doolen (1998); Guo, Zheng & Shi (2002); Rohde et al. (2006) | de Vahl Davis Ra = 1e3–1e6 with refined walls; Tric et al. (2000) |
| 2 | Thermal BCs (heat flux, Robin, interpolated curved) | high | low | low | high | M | Li, Mei & Klausner (2013, JCP 237:366) | concentric-annulus natural convection (Kuehn & Goldstein 1976); heated cylinder in a channel |
| 3 | Thermal collision stable at high Ra, plus thermal LES (MRT / regularised, Pr_t, Kader validated) | high | low | low | high | M | Contrino et al. (2014, JCP 275:257); Kader (1981) | Le Quéré (1991) Ra = 1e8; Tian & Karayiannis (2000) |
| 4 | Conjugate heat transfer | mid | low | low | high | M | Karani & Huber (2015, PRE 91:023304) | Kaminski & Prakash (1986) cavity with a conducting wall |
| 5 | Production dynamic AMR (validated regrid, sensor gates) | high | high | mid | mid | L | Schornbaum & Rüde (2018); Eitel-Amor et al. (2013) | cylinder Re = 100 with regrid vs uniform (St, Cd); advected Lamb–Oseen vortex |
| 6 | Deep-level interface accuracy (trustworthy forces, D3Q19 divergence) | mid | high | low | low | L | Astoul et al. (2020); Rohde et al. (2006); Dorschner et al. (2016) | sphere Cd over levels 1–4; sound crossing the interface |
| 7 | Dynamic load balancing with MPI migration | mid | high | mid | mid | L | Schornbaum & Rüde (2018); Burstedde et al. (2011) | weak scaling with dynamic AMR; bitwise match to one rank |
| 8 | Humidity, age of air and other scalars on the forest | high | low | low | high | S | Sandberg (1981) (age of air) | IEA Annex 20 (Nielsen 1990) with a tracer |
| 9 | Wall model that works with separation (WMLES) | mid | high | low | mid | L | Malaspinas & Sagaut (2014); Haussmann et al. (2019) | Lee & Moser (2015) Re_τ = 1000; periodic hill; Ahmed slant Cp |
| 10 | Convective outflow and non-reflecting BC / sponge | high | high | mid | mid | S–M | Yang (2013); Izquierdo & Fueyo (2008); Xu & Sagaut (2013) | vortex reflection at the outlet; acoustic reflection coefficient |
| 11 | 3D radiation coupled to temperature | mid | low | low | high | M | Asinari et al. (2010); Mishra & Roy (2007); Modest (2013) | Akiyama & Chong (1997) radiation with natural convection |
| 12 | Free-surface PLIC and surface tension, impact stability | low | low | high | low | M–L | Sato & Koshimura (2020); Lehmann & Gekle (2022) | SPHERIC test 2 P1; interface advection in a deformation field |
| 13 | Free-surface inflow/outflow and wave making | low | low | high | low | M | Stoker (1957) analytic; Synolakis (1987) | dam break on a wet bed; solitary-wave run-up |
| 14 | Free surface on the forest AMR | low | low | high | low | L | Thürey & Rüde (2009); Schwarzmeier et al. (2023) | Koshizuka & Oka against a uniform grid |
| 15 | IBM for moving geometry (fans, louvres, wheels) | mid | mid | low | mid | M | Uhlmann (2005); Feng & Michaelides (2004) | oscillating cylinder (Dütsch et al. 1998) |
| 16 | HDF5 and layout-independent checkpoints | mid | mid | mid | mid | S–M | HDF5 specification | bitwise match after restart |
| 17 | In-situ visualisation | mid | mid | mid | mid | M | Ayachit et al. (2015); Larsen et al. (2017) | image-output regression test |
| 18 | AMD / Intel GPUs (Kokkos) | mid | mid | mid | mid | L | Trott et al. (2022) | bitwise match to stdpar (cavity, Couette) |
| 19 | Vreman SGS | mid | mid | low | mid | S | Vreman (2004) | TGV Re = 1600; channel |
| 20 | PSM (partially saturated) | low | low | low | low | M | Noble & Torczynski (1998) | sphere drag on a coarse grid; Ten Cate |
| 21 | Shan–Chen / free energy | low | low | low | low | S–M | Shan & Chen (1993); Swift et al. (1996) | Laplace pressure; contact angle |
| 22 | Code generation and a Python front end | low | low | low | low | L | Bauer et al. (2021) (ideas only) | bitwise match to hand-written kernels |
Missing validation cases (ranked)¶
- Heated cavity at Ra = 1e6 (de Vahl Davis 1983) and Ra = 1e8 (Le Quéré 1991).
- 3D heated cube (Fusegi et al. 1991, Tric et al. 2000), with the reference values recorded in the repository.
- Mixed convection in a ventilated room (Blay, Mergui & Niculae 1992).
- Room ventilation (IEA Annex 20, Nielsen 1990).
- Curved thermal walls: concentric-annulus natural convection (Kuehn & Goldstein 1976), heated cylinder in a channel.
- Conjugate heat transfer cavity (Kaminski & Prakash 1986).
- Turbulent natural convection (Tian & Karayiannis 2000).
- 3D Rayleigh–Bénard (Nu against Ra).
- AMR: drag convergence of refined spheres and cylinders against a uniform grid; quantitative vortex and sound crossing an interface (Astoul et al. 2020).
- 3D cylinder (Schäfer & Turek 3D-1Z / 3D-2Z).
- Wall-resolved channel (Moser, Kim & Mansour 1999) and wall-modelled channel (Lee & Moser 2015).
- Periodic hill (Breuer et al. 2009).
- Free surface: dam break on a wet bed (Stoker), dam-break impact pressure (Lobovsky et al. 2014), interface advection in a deformation field.
- Oscillating cylinder (Dütsch et al. 1998).
- 3D cavity (Albensoeder & Kuhlmann 2005).
- Permeability (sphere arrays, Zick & Homsy 1982).
- Fluidised bed, settling cube.
The Ahmed body and AIJ Case A have cases but lack recorded comparison numbers; see Validation cases.
Sources¶
The public material used for the other codes. No source code.
- W1 — waLBerla home page (features). https://www.walberla.net/
- W2 — Bauer et al. (2021), waLBerla: a block-structured high-performance framework for multiphysics simulations, Comput. Math. Appl. 81:478–501. https://arxiv.org/abs/1909.13772
- W3 — Bauer et al. (2021), lbmpy: automatic code generation for efficient parallel lattice Boltzmann methods, Comput. Phys. Commun. 259:107522; lbmpy documentation. https://pycodegen.pages.i10git.cs.fau.de/lbmpy/
- W4 — lbmpy documentation, Boundary Conditions. https://pycodegen.pages.i10git.cs.fau.de/lbmpy/sphinx/boundary_conditions.html
- W5 — Schornbaum & Rüde (2016), Massively parallel algorithms for the lattice Boltzmann method on nonuniform grids, SIAM J. Sci. Comput. 38(2):C96–C126.
- W6 — Schornbaum & Rüde (2018), Extreme-scale block-structured adaptive mesh refinement, SIAM J. Sci. Comput. 40(3):C358–C387.
- W7 — Rettinger & Rüde (2017), A comparative study of fluid-particle coupling methods for fully resolved lattice Boltzmann simulations, Comput. Fluids 154:74–89.
- W8 — Holzer et al. (2021), Highly efficient lattice Boltzmann multiphase simulations of immiscible fluids at high-density ratios on CPUs and GPUs through code generation, Int. J. High Perform. Comput. Appl. 35(4):413–427.
- W9 — Schwarzmeier et al. (2023), Comparison of free-surface and conservative Allen–Cahn phase-field lattice Boltzmann method, J. Comput. Phys. 473:111753.
- W10 — waLBerla repository, names of the apps (benchmarks, showcases, tutorials) only. https://i10git.cs.fau.de/walberla/walberla/-/tree/master/apps
- O1 — Krause et al., OpenLB User Guide (release 1.6). https://arxiv.org/abs/2307.11752
- O2 — OpenLB release 1.8 announcement. https://www.openlb.net/news/openlb-release-1-8-available-for-download/
- O3 — OpenLB release 1.9 announcement. https://www.openlb.net/news/openlb-release-1-9-available-for-download/
- O4 — OpenLB 1.9.0 release archive, names of the example directories only. https://zenodo.org/records/17899765
- O5 — Krause et al. (2021), OpenLB — open source lattice Boltzmann code, Comput. Math. Appl. 81:258–288.
- G1 — Lagrava, Malaspinas, Latt & Chopard (2012), Advances in multi-domain lattice Boltzmann grid refinement, J. Comput. Phys. 231:4808–4822.
- P1 — Latt et al. (2021), Palabos: parallel lattice Boltzmann solver, Comput. Math. Appl. 81:334–350.
- P2 — Latt et al. (2025), Multi-GPU acceleration of PALABOS fluid solver using C++ standard parallelism. https://arxiv.org/abs/2506.09242
- P3 — Palabos repository, names of the example directories only. https://gitlab.com/unigespc/palabos/-/tree/master/examples
- P4 — Palabos home page. https://palabos.unige.ch/
- F1 — FluidX3D README (feature list). https://github.com/ProjectPhysX/FluidX3D
- F2 — Lehmann et al. (2022), Accuracy and performance of the lattice Boltzmann method with 64-bit, 32-bit, and customized 16-bit number formats, Phys. Rev. E 106:015308. https://arxiv.org/abs/2112.08926
- F3 — Lehmann (2022), Combined scientific CFD simulation and interactive raytracing with OpenCL, IWOCL '22.
- F4 — Lehmann & Gekle (2022), Analytic solution to the piecewise linear interface construction problem and its application in curvature calculation for volume-of-fluid simulation codes, Computation 10(2):21.
- T1 — TCLB documentation (model list). https://docs.tclb.io/
- T2 — Łaniewski-Wołłk & Rokicki (2016), Adjoint lattice Boltzmann for topology optimization on multi-GPU architecture, Comput. Math. Appl. 71:833–848.
Related pages: Cross-code comparison, Architecture, AMR, Turbulence and walls, Free surface, Known issues.