Setup Guide: Laminar-Turbulent Transition

Draft. These settings are drawn from SU2 source, tutorials and regression cases. Recommendations marked (to be reviewed) still need an experienced user’s review.

Use this guide to choose settings and understand their consequences. Follow the linked Tutorials for worked examples and consult V&V for numerical and physical comparisons. Example settings are starting points, not evidence that a case is converged or validated.

The documentation guide explains the source tags used below: (code), (test case), (tutorial), (docs), (advice) and (to be reviewed).

When to use this guide

A standard RANS run treats the boundary layer as turbulent from the leading edge. A transition model lets the boundary layer start laminar and become turbulent further downstream. The model damps the production of turbulence until a transition criterion is met (tutorial).

Use a transition model when laminar flow covers a large part of the surface, for example (advice):

  • airfoils and blades at Reynolds numbers from about 10^5 to a few 10^6,
  • wind turbine blades, gliders, natural laminar flow wings,
  • cases with laminar separation bubbles.

A fully turbulent model cannot represent the laminar region and its transition mechanisms. It can overpredict local skin friction there, but the change in total drag also depends on separation and pressure drag (advice).

Limits of the Langtry-Menter (LM) model in SU2 (tutorial):

  • It is built from subsonic experiments: transition caused by Tollmien-Schlichting waves, by free-stream turbulence (“bypass” transition) and by laminar separation.
  • It does not predict crossflow transition (except through the LM2015 option, see below), nor supersonic and hypersonic transition mechanisms.

LM has no discrete adjoint in this code version. Do not use KIND_TRANS_MODEL= LM with MATH_PROBLEM= DISCRETE_ADJOINT; see Transition Models (code).

For the RANS settings that are not specific to transition, see Setup Guide: Steady RANS for External Aerodynamics.

Start from

This config comes from the E387 airfoil tutorial with SA and LM: compressible_flow/Transitional_Airfoil/Langtry_and_Menter/E387/transitional_SA_LM_model_ConfigFile.cfg in the Tutorials repository. The SU2 tutorial regression tests run it. We removed the output file name options and JST_SENSOR_COEFF (it has no effect with ROE). All other values are unchanged.

% ------------- Problem ------------------------------------------------%
SOLVER= RANS
KIND_TURB_MODEL= SA
KIND_TRANS_MODEL= LM
FREESTREAM_TURBULENCEINTENSITY= 0.001
MATH_PROBLEM= DIRECT
RESTART_SOL= NO
%
% ------------- Free stream --------------------------------------------%
MACH_NUMBER= 0.06
AOA= 2
FREESTREAM_TEMPERATURE= 288.15
REYNOLDS_NUMBER= 200000.0
REYNOLDS_LENGTH= 1.0
REF_ORIGIN_MOMENT_X= 0.25
REF_ORIGIN_MOMENT_Y= 0.00
REF_ORIGIN_MOMENT_Z= 0.00
REF_LENGTH= 1.0
REF_AREA= 0
REF_DIMENSIONALIZATION= FREESTREAM_VEL_EQ_MACH
%
% ------------- Boundary conditions ------------------------------------%
MARKER_HEATFLUX= ( Airfoil, 0.0 )
MARKER_FAR= ( Farfield )
MARKER_PLOTTING= ( Airfoil )
MARKER_MONITORING= ( Airfoil )
%
% ------------- Numerics -----------------------------------------------%
NUM_METHOD_GRAD= GREEN_GAUSS
CFL_NUMBER= 1
CFL_ADAPT= YES
CFL_ADAPT_PARAM= ( 0.1, 2.0, 40.0, 1e10, 0.001 )
ITER= 100000
LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= LU_SGS
LINEAR_SOLVER_ERROR= 1E-6
LINEAR_SOLVER_ITER= 5
CONV_NUM_METHOD_FLOW= ROE
USE_VECTORIZATION= NO
MUSCL_FLOW= YES
SLOPE_LIMITER_FLOW= VENKATAKRISHNAN
VENKAT_LIMITER_COEFF= 0.05
LIMITER_ITER= 99999
TIME_DISCRE_FLOW= EULER_IMPLICIT
CONV_NUM_METHOD_TURB= SCALAR_UPWIND
MUSCL_TURB= NO
SLOPE_LIMITER_TURB= NONE
TIME_DISCRE_TURB= EULER_IMPLICIT
CFL_REDUCTION_TURB= 1.0
%
% ------------- Convergence --------------------------------------------%
CONV_FIELD= ( RMS_DENSITY )
CONV_RESIDUAL_MINVAL= -13
CONV_STARTITER= 500
%
% ------------- Input and output ---------------------------------------%
MESH_FILENAME= ogrid_e387.su2
MESH_FORMAT= SU2
OUTPUT_FILES= (RESTART, PARAVIEW, SURFACE_PARAVIEW)
WRT_FORCES_BREAKDOWN= YES
VOLUME_OUTPUT= (COORDINATES, SOLUTION, PRIMITIVE, RESIDUAL)
HISTORY_OUTPUT= (ITER, LINSOL, RMS_RES, MAX_RES, AERO_COEFF, CFL_NUMBER)
OUTPUT_WRT_FREQ= 500
SCREEN_WRT_FREQ_INNER= 10
SCREEN_OUTPUT= (INNER_ITER, WALL_TIME, RMS_DENSITY, RMS_INTERMITTENCY, RMS_RE_THETA_T, LIFT, DRAG, LINSOL_ITER_TRANS, LINSOL_RESIDUAL_TRANS)
WRT_PERFORMANCE= YES

The same tutorial folder has an SST version, transitional_SST_LM_model_ConfigFile.cfg. It differs in three lines (tutorial):

KIND_TURB_MODEL= SST
SST_OPTIONS= V1994m
MUSCL_TURB= YES

Settings that matter

Turn on the LM model

  • KIND_TRANS_MODEL= LM turns on the Langtry-Menter model. NONE (the default) and LM are the only values (code).
  • LM needs a turbulence model. With KIND_TURB_MODEL= NONE, SU2 stops with an error (code).
  • LM adds two transport equations (code):
    • the intermittency, gamma: 0 where the boundary layer is laminar, 1 where it is turbulent;
    • the transition onset Reynolds number, Re_theta_t: a Reynolds number based on the momentum thickness, at which transition starts.
  • The model uses gamma to damp the production term of the turbulence model (tutorial).

Base turbulence model: SST or SA

The develop branch of SU2 supports LM with both SST and SA (code, docs). The two couplings differ in how they get the turbulence intensity Tu. Tu drives the transition correlations, the empirical formulas that predict where transition starts:

  KIND_TURB_MODEL= SST KIND_TURB_MODEL= SA
Turbulence intensity used by LM Local value from k: Tu = 100·sqrt(2k/3)/U, not below 0.027 % (code) FREESTREAM_TURBULENCEINTENSITY everywhere (code)
Default correlation MENTER_LANGTRY (code) MALAN (code)
Tutorials T3A and T3A- flat plates, E387 airfoil (tutorial) E387 airfoil (tutorial)

What this means (advice, to be reviewed):

  • With SST, the intensity at the body depends on how much k decays on its way from the far field. You control it with FREESTREAM_TURBULENCEINTENSITY and FREESTREAM_TURB2LAMVISCRATIO.
  • With SA, LM uses the same intensity everywhere, so nothing decays. Set FREESTREAM_TURBULENCEINTENSITY to the intensity you expect at the leading edge.

The original LM model was published with SST; see the TMR model definition (reference).

The supported variants here are those registered in this source revision; check the generated reference before using options from another branch.

LM_OPTIONS

LM_OPTIONS takes a list. With an empty list, SU2 runs the 2009 version of the model with the default correlation of the base turbulence model (code).

Version.

  • LM2015 adds a correction for stationary crossflow instabilities (docs). It works in 3D only; in 2D SU2 stops with an error (code).
  • LM2015 uses the surface roughness HROUGHNESS (RMS roughness height, default 1e-6) (code, docs). The code divides it by a momentum thickness, so give it in the same length unit as the mesh (code, to be reviewed).

Correlation. The correlation computes the transition onset Reynolds number. Choose at most one; two or more stop SU2 with an error (code).

Value Notes
MENTER_LANGTRY Default with SST (code).
MALAN Default with SA (code, docs).
SULUKSNA Use with SST only (docs).
KRAUSE, KRAUSE_HYPER Meant for hypersonic flows. The docs say their implementation is unclear because of inconsistencies in the literature (docs).
MEDIDA_BAEDER Designed for SA. Has problems with separation-induced transition (docs).
DEFAULT Same as leaving the correlation out (code).

Some pages also list MEDIDA. The develop code does not accept it as a keyword (code).

Keep the default correlation unless you have a reason to change it, for example a validation for your type of flow (advice).

Example: LM_OPTIONS= LM2015, MENTER_LANGTRY (docs).

Free-stream turbulence intensity and viscosity ratio

The free-stream turbulence intensity is the main input of the transition correlations (tutorial).

  • FREESTREAM_TURBULENCEINTENSITY is a fraction, not a percentage: 0.001 means 0.1 %. SU2 multiplies it by 100 inside the correlations (code).
  • The default is 0.05 (5 %) (code). This is high for most external flows (docs).
  • SU2 computes the far-field value of Re_theta_t from this intensity with the Langtry-Menter correlation (code).
  • FREESTREAM_INTERMITTENCY has no effect on LM: SU2 prints it, but the far-field intermittency is fixed at 1 (code).

With SST.

  • SU2 builds the far-field k from the intensity, and omega from FREESTREAM_TURB2LAMVISCRATIO (eddy viscosity over laminar viscosity, default 10.0) (code).
  • k decays between the far field and the body. A larger viscosity ratio gives a slower decay (advice, to be reviewed).
  • The T3A tutorial sets FREESTREAM_TURBULENCEINTENSITY= 0.05855 and FREESTREAM_TURB2LAMVISCRATIO= 11.9 at the inflow (tutorial).
  • Check the intensity just upstream of the body. Compute it from the TKE output field: Tu = sqrt(2k/3)/U (advice).
  • SST_OPTIONS= SUSTAINING adds ambient sustaining terms through a production floor in SU2. TURB_FIXED_VALUES= YES with TURB_FIXED_VALUES_DOMAIN fixes the turbulence variables in a specified upstream region (code). These change the turbulence environment; match the measured intensity and decay rather than removing decay automatically (advice). No transition test case uses them with LM (to be reviewed).

With SA.

  • FREESTREAM_TURB2LAMVISCRATIO is not used. The free-stream SA variable comes from FREESTREAM_NU_FACTOR (default 3.0) (code, docs).
  • The E387 tutorial uses FREESTREAM_TURBULENCEINTENSITY= 0.001 (0.1 %) (tutorial).

Mesh

General points for the mesh of an LM run (advice, to be reviewed):

  • First-cell y+ about 1 or below, all along the walls. Do not use wall functions.
  • A smooth wall-normal growth in the boundary layer, with a growth ratio near 1.1.
  • Enough points along the wall where transition or a laminar separation bubble is expected. The transition location moves on meshes that are too coarse in this direction.
  • Check the transition location on at least two meshes. Transition models are more sensitive to the mesh than fully turbulent RANS.

Numerics

  • The transition equations use the settings of the turbulence equations: CONV_NUM_METHOD_TURB, MUSCL_TURB, SLOPE_LIMITER_TURB and TIME_DISCRE_TURB (code).
  • The tutorials use ROE with MUSCL and VENKATAKRISHNAN (E387), and L2ROE with MUSCL (T3A) for the flow (tutorial). L2ROE is a Roe scheme with low dissipation at low Mach number (docs).
  • Both E387 configs start at CFL_NUMBER= 1 with CFL_ADAPT= YES and use LU_SGS (tutorial).
  • Transition cases often run at low Mach numbers (0.06 for E387, 0.2 for T3A) (tutorial). The compressible solver converges more slowly at very low Mach numbers (advice, to be reviewed).
  • Add RMS_INTERMITTENCY, RMS_RE_THETA_T, LINSOL_ITER_TRANS and LINSOL_RESIDUAL_TRANS to SCREEN_OUTPUT to follow the transition equations (tutorial, code).

The algebraic BC model: a lighter option with SA

The develop branch also has an algebraic transition model for SA, the Bas-Cakmakcioglu (BC) model: KIND_TURB_MODEL= SA with SA_OPTIONS= BCM (tutorial, code).

  • It adds no transport equation. It damps the SA production term until a local criterion is met (tutorial).
  • It also uses FREESTREAM_TURBULENCEINTENSITY as a fraction (code). The tutorial uses 0.0018 for 0.18 % (tutorial).
  • Do not combine BCM with other SA options. SU2 stops with an error unless you add EXPERIMENTAL (code).
  • The old option KIND_TRANS_MODEL= BC is removed. SU2 stops with a message that points to SA_OPTIONS= BCM (code).
  • The SA-BCM INTERMITTENCY output currently returns zero instead of the computed intermittency. Do not use it to locate transition; inspect the skin friction and eddy viscosity instead (code, advice). See Transition Models.
  • See the tutorial Transitional Flat Plate.

Check convergence and results

Convergence.

  • Residuals: RMS_DENSITY, the turbulence residuals, RMS_INTERMITTENCY and RMS_RE_THETA_T (code).
  • Lift and drag must be flat at the end (advice).
  • The transition location must stop moving. Compare the skin friction plots of the last saved solutions (advice).

Where transition happens. The default output groups SOLUTION and PRIMITIVE contain the fields below (code).

Field Group What to look for
SKIN_FRICTION-X (surface) PRIMITIVE Low in the laminar part, then a sharp rise at transition. For flow aligned with +x, negative values indicate reversed wall shear, which can occur in either laminar or turbulent separation. Use the local streamwise shear component on a curved or inclined surface (advice).
INTERMITTENCY SOLUTION Transported intermittency: 0 laminar, 1 turbulent (code).
INTERMITTENCY_EFF PRIMITIVE Effective intermittency that multiplies the production term. It can exceed 1 in separated regions (advice, to be reviewed).
INTERMITTENCY_SEP PRIMITIVE Extra intermittency for separation-induced transition (advice, to be reviewed).
RE_THETA_T SOLUTION Transported transition onset Reynolds number (code).
EDDY_VISCOSITY PRIMITIVE Rises where the boundary layer becomes turbulent (advice).
TURB_INDEX PRIMITIVE Turbulence index on wall nodes: near 0 laminar, near 1 turbulent. Use it only as a rough indicator (code, to be reviewed).

Plot SKIN_FRICTION-X from the surface output (SURFACE_PARAVIEW or SURFACE_CSV in OUTPUT_FILES) along the wall, and compare it with experiments. The T3A tutorial compares it with ERCOFTAC data as a function of the local Reynolds number (tutorial).

See the Output Fields Reference.

Common problems

Symptom Likely cause What to try
The boundary layer is turbulent from the leading edge. Turbulence intensity too high; or given in percent (for example 3.0 instead of 0.03). Give FREESTREAM_TURBULENCEINTENSITY as a fraction (code). Check the intensity near the body (advice).
Transition happens too late, or not at all. With SST, k has decayed to a very low value before the body. Or the mesh is too coarse at the wall. Check Tu near the leading edge from TKE; adjust FREESTREAM_TURB2LAMVISCRATIO (advice, to be reviewed). Check y+ (advice).
The transition location moves when you refine the mesh. Too few points along the wall in the transition region. Refine along the wall (advice).
The residuals stall or oscillate. A laminar separation bubble that is unsteady, or a CFL that is too high. Start at a low CFL with CFL_ADAPT= YES (tutorial). If the bubble is unsteady, run time-accurate, see Setup Guide: Unsteady Flows and DES (advice, to be reviewed).
SU2 stops: “LM2015 is available only for 3D problems”. LM2015 in a 2D case. Remove LM2015 (code).
SU2 stops: “Two correlations selected for LM_OPTIONS”. More than one correlation in LM_OPTIONS. Keep one (code).
SU2 stops: “KIND_TURB_MODEL cannot be NONE to use a transition model”. No base turbulence model. Set KIND_TURB_MODEL= SST or SA (code).
SU2 stops: “Option KIND_TRANS_MODEL=BC is deprecated”. Old config. Use KIND_TURB_MODEL= SA with SA_OPTIONS= BCM (code).
Changing FREESTREAM_INTERMITTENCY does nothing. The option is not used by the solver. Leave it out (code).
An old config stops with “invalid option name” for EXT_ITER. EXT_ITER was replaced. Use ITER for steady runs (code).

Tutorials:

Test cases (develop branch):

User guide pages:

Source review: 22 October 2026, against SU2 docs_site at 41ea6a219a. See the review scope and outstanding validation. Setup recommendations remain draft where indicated.


Improve this page