Numerical Schemes
The Numerical section controls the spatial discretization of convective and viscous fluxes and provides parameters to stabilize high-speed or shock-dominated flows.
Overview
This block specifies:
- conv_scheme – discretization of the convective terms
- conv_scheme_ibm – convective scheme used at immersed boundary (IBM) surfaces
- visc_scheme – viscous flux discretization
- Stabilization – tuning parameters for hybrid convective schemes
Convective Scheme (conv_scheme)
Convective schemes fall into two broad categories:
- Ordinary schemes – a single discretization (e.g. pure WENO or MUSCL2)
- Hybrid schemes – fourth-order central differencing in smooth regions, blended automatically with upwind WENO fluxes near shocks or turbulence using a dilatation-based shock sensor
For ideal_gas the available schemes are:
| Scheme | Type | Notes / Recommended use |
|---|---|---|
central4_charweno | Hybrid | For turbulence and strong shocks. Fourth-order central base with characteristic WENO3 upwinding. Recommended option for supersonic LES. |
central4_charweno_lowdiss | Hybrid | As above but with a lower-dissipation, non-conservative flux. For LES or very low Mach numbers. |
central4_weno3fvs | Hybrid | Uses WENO3 with flux-vector splitting; robust and stable, slightly higher dissipation than FDS. Recommended option for subsonic LES. |
central4_weno3fds | Hybrid | Uses WENO3 with flux-difference splitting; slightly lower dissipation, slightly less robust. |
central4_weno3fds_hllc | Hybrid | Variant using an HLLC flux; niche option for extreme shocks. |
central4_weno3fvs_lowdiss | Hybrid | Non-conservative, lower-dissipation FVS variant. |
central4_weno3fds_lowdiss | Hybrid | Non-conservative, lower-dissipation FDS variant. |
central4 | Ordinary | Pure fourth-order central, kinetic/entropy preserving. Mostly for research. |
weno3 | Ordinary | Pure 3rd-order WENO (no central blending). Robust and moderately dissipative. |
charweno | Ordinary | Pure 3rd-order characteristic WENO; handles strong shocks, higher dissipation. Recommended option for inviscid hypersonic simulations. |
charweno_ausm | Ordinary | Characteristic WENO with AUSM flux. Good for weak shocks and jets, less robust on strong shocks. |
muscl2 | Ordinary | Classical 2nd-order MUSCL. Legacy or low-cost runs only. |
first_order | Ordinary | Very diffusive; emergency-only for debugging or pathological cases. |
none | Ordinary | Disables convection entirely. For manufactured-solution or special test cases only. |
Other physical models (e.g. air_CO2, air_SF6, air5, air11) may support different sets of convective schemes.
These will be documented in future updates.
Convective Scheme near IBM (conv_scheme_ibm)
This controls the discretization near immersed boundaries:
auto(default): mimics the interior scheme while maintaining stability with cut-cell geometry.first_order: forces first-order upwind near IBM boundaries for maximum robustness (more diffusive).
Viscous Scheme (visc_scheme)
Options:
visc2(recommended): standard second-order viscous discretization.none: disables viscous fluxes, suitable only for inviscid (Euler) test problems.
Stabilization
Hybrid convective schemes introduce additional stabilization controls:
| Parameter | Meaning | Guidance |
|---|---|---|
ducr_shield | Physical distance outside which the hybrid sensor is set to 1.0 (fully upwind). | Normally leave unset. |
ducr_multiplier | Scales the dilatation-based shock sensor. | Usually 1.0. |
ducr_cutoff | If the dilatation sensor is below this value, it is set to zero. | Lowering increases dissipation. |
minimum_value | Minimum blending value anywhere in the flow. | Low Mach: ~1e-3–5e-3. Hypersonic: 0.01–0.1. Higher = more stable. |
minimum_value_ibm | Minimum blending value applied at immersed boundaries. | Set similar to minimum_value. |
These parameters allow tuning for challenging cases such as very low Mach numbers or hypersonic flows with strong shocks.
Reflux
reflux (default: false)
Enables conservative flux correction across coarse–fine grid interfaces (AMR).
When true, the solver ensures that net fluxes are exactly conserved between adjacent blocks of different resolution.
This slightly increases computational cost and memory but improves global conservation of mass, momentum, and energy, especially for long-time or high-accuracy runs.
PrototypeRefluxing on multi-GPU simulations is currently not supported. It's not recommended to use refluxing.
Example
Numerical
{
conv_scheme = "central4_charweno"
conv_scheme_ibm = "auto"
visc_scheme = "visc2"
reflux = false
Stabilization
{
ducr_shield = 1.0
ducr_multiplier = 1.0
ducr_cutoff = 0.25
minimum_value = 0.01
minimum_value_ibm = 0.01
}
}
This configuration
- uses a fourth-order central + characteristic WENO hybrid scheme for the convective fluxes,
- sets the immersed-boundary scheme to automatically match the interior scheme, and
- applies second-order viscous discretization.