8.4 Boundary conditions

Types of boundary conditions

Boundary conditions define how the flow behaves at the edges of your computational domain. Without them, the governing PDEs have infinitely many solutions. By specifying what happens at each boundary, you constrain the problem so it has a unique, physically meaningful answer.

Different boundary condition types impose constraints on either the flow variables themselves or their derivatives at the boundaries.

Dirichlet vs Neumann

Dirichlet boundary conditions fix the value of a variable directly at the boundary. You're telling the solver "the velocity here is exactly this" or "the temperature at this wall is 300 K." Common uses include:

• Inflow boundaries (prescribed velocity profile)
• No-slip walls (velocity = 0 relative to the wall)
• Prescribed temperature surfaces

Neumann boundary conditions fix the normal derivative of a variable at the boundary. Instead of specifying the value, you're specifying how fast it changes in the direction perpendicular to the boundary. Common uses include:

• Outflow boundaries (zero velocity gradient, so the flow exits undisturbed)
• Symmetry planes (zero normal gradient by definition)
• Adiabatic walls (zero heat flux means $\frac{\partial T}{\partial n} = 0$)

Robin boundary conditions

Robin boundary conditions combine Dirichlet and Neumann into a single linear relationship between a variable's value and its normal derivative at the boundary. Two classic examples:

• Convective heat transfer: $q = h(T_s - T_\infty)$, where the heat flux depends on the difference between the surface temperature and the freestream temperature
• Slip walls: $u_s = \lambda \frac{\partial u}{\partial n}$, where the wall velocity is proportional to the velocity gradient (relevant in rarefied gas dynamics)

These are more physically realistic than pure Dirichlet or Neumann conditions in many heat transfer and microflow applications.

Mixed boundary conditions

Mixed boundary conditions apply different condition types to different variables or different parts of the same boundary. For example, a heated no-slip wall uses Dirichlet for both velocity (zero at the wall) and temperature (prescribed wall temperature), but you could also have Dirichlet for velocity and Neumann for temperature (adiabatic wall) on the same surface. This flexibility lets you build more realistic representations of complex physical boundaries.

Physical significance

Boundary conditions aren't just mathematical requirements. They represent real physical constraints: how a fluid interacts with a wall, what happens far from the body, and what the flow looks like entering or leaving the domain.

Flow field constraints

• Inflow boundaries: You prescribe the velocity profile, turbulence quantities, and thermodynamic state (pressure, temperature). These define the flow entering your domain.
• Outflow boundaries: Typically zero-gradient or extrapolation conditions that let the flow exit smoothly without artificially influencing the upstream solution.
• No-slip walls: Velocity is zero relative to the wall surface. This is the condition that creates the viscous boundary layer in the first place.

Surface interactions

• Isothermal walls: Surface temperature is fixed at a known value. Used when the wall material has high thermal conductivity or active cooling.
• Adiabatic walls: Zero heat flux through the surface (perfectly insulated). The wall temperature adjusts to match the flow.
• Slip walls: Non-zero tangential velocity at the wall. Applied in inviscid simulations or rarefied gas flows where the mean free path is comparable to the flow length scale.
• Catalytic walls: Surface chemical reactions (adsorption, desorption, recombination) occur at the boundary. Relevant for hypersonic reentry vehicles where dissociated gas species interact with the thermal protection system.

Far-field behavior

Far-field boundaries model the "infinity" surrounding an external aerodynamics problem. Since you can't make the domain infinitely large, you place boundaries far enough away and apply conditions that minimize their influence on the solution near the body.

• Freestream conditions: Directly impose undisturbed flow properties.
• Characteristic-based conditions: Use the mathematical characteristics of the governing equations to determine which variables should be specified and which should be extrapolated, reducing spurious wave reflections.
• Infinite elements: Map the far-field region to infinity, eliminating truncation effects entirely.

Mathematical formulation

Boundary conditions supplement the governing PDEs to ensure the problem is well-posed, meaning a unique solution exists and depends continuously on the input data. The type and number of boundary conditions you need depend on the order and classification of the PDEs you're solving.

Partial differential equations

Boundary conditions accompany PDEs like the Navier-Stokes equations, the heat equation, and the wave equation. The classification of the PDE determines what boundary conditions are appropriate:

• Elliptic PDEs (e.g., Laplace equation for steady potential flow) need conditions on the entire boundary of the domain.
• Parabolic PDEs (e.g., unsteady heat equation) need conditions on the spatial boundaries plus an initial condition.
• Hyperbolic PDEs (e.g., wave equation, Euler equations) need conditions based on characteristic directions, and specifying too many or too few causes the problem to be ill-posed.

For the incompressible Navier-Stokes equations, you need boundary conditions for both velocity and pressure.

Initial value problems

Initial value problems (IVPs) require the flow field to be specified at a starting time $t = t_0$. This gives the solver a known state to march forward from. Spatial boundary conditions are still required at every time step. Example: an unsteady simulation of vortex shedding behind a cylinder, where you prescribe the initial velocity and pressure fields throughout the domain.

Boundary value problems

Boundary value problems (BVPs) require conditions only on the spatial boundaries, with no time dependence. Elliptic PDEs like the Laplace equation for steady heat conduction or potential flow are classic BVPs. A typical example: potential flow around an airfoil, where you impose a no-penetration condition on the airfoil surface and prescribe the freestream velocity at the far-field boundary.

Numerical implementation

Even if your boundary conditions are physically correct, poor numerical implementation can introduce errors, instabilities, or non-physical solutions. The treatment depends on your discretization method.

Finite difference methods

1. Modify the finite difference stencils near boundaries where the standard stencil would reach outside the domain.
2. Use ghost cells (fictitious cells outside the domain whose values are set to enforce the boundary condition) or one-sided differences to maintain accuracy.
3. Dirichlet conditions are straightforward: set the boundary node value directly. Neumann conditions require approximating the derivative, typically with a one-sided finite difference.

Finite element methods

1. Rewrite the governing equations in weak form by multiplying by a test function and integrating over the domain.
2. Dirichlet conditions are enforced by restricting the trial function space so that the solution takes the prescribed values on the boundary.
3. Neumann conditions appear naturally in the boundary integral terms of the weak form. If you don't explicitly specify anything on a boundary, the finite element method automatically imposes a zero-flux (homogeneous Neumann) condition. This is why Neumann conditions are sometimes called natural boundary conditions in the FEM context.

Spectral methods

1. Choose basis functions (Chebyshev polynomials, Fourier series) that can satisfy the boundary conditions.
2. Dirichlet conditions can be built directly into the basis functions so they're satisfied by construction.
3. Neumann conditions are typically imposed using the tau method (adding extra equations that enforce the derivative condition) or the penalty method (adding a large penalty term that drives the solution toward satisfying the condition).

Boundary layer theory

The boundary layer is the thin region near a surface where viscous effects dominate. Boundary conditions at the wall and at the outer edge of this layer control its entire structure, including whether it stays laminar, transitions to turbulence, or separates.

Viscous effects near boundaries

The no-slip condition forces the fluid velocity to match the wall velocity (usually zero). This creates steep velocity gradients near the surface, and those gradients produce shear stress and viscous dissipation. Surface friction drag and convective heat transfer are both governed by conditions within this thin layer.

Boundary layer equations

The boundary layer equations are a simplified form of the Navier-Stokes equations, derived by assuming the layer is thin relative to the body length ($\delta \ll L$). This scale analysis eliminates the pressure gradient normal to the wall and reduces the equations to a more tractable form. They require:

• Wall boundary conditions: no-slip ($u = 0$ at $y = 0$) and a thermal condition (prescribed temperature or heat flux)
• Outer edge condition: velocity matches the freestream or inviscid solution ($u \to U_e$ as $y \to \delta$)

Laminar vs turbulent boundary layers

• Laminar boundary layers have smooth, ordered streamlines with momentum transport driven only by molecular viscosity.
• Turbulent boundary layers exhibit chaotic, unsteady fluctuations that dramatically increase mixing, heat transfer, and skin friction compared to laminar flow.
• Transition from laminar to turbulent depends on the Reynolds number, pressure gradient, surface roughness, and freestream turbulence.
• Turbulent boundary layers require additional modeling (turbulence models, wall functions) and their own set of boundary conditions for turbulence quantities like $k$ and $\omega$.

Shock wave boundary conditions

Shock waves are extremely thin regions where flow properties change abruptly. Across a shock, pressure, density, temperature, and velocity all jump to new values. The boundary conditions governing these jumps come from conservation laws applied across the discontinuity.

Rankine-Hugoniot relations

The Rankine-Hugoniot relations enforce conservation of mass, momentum, and energy across a shock wave. They relate upstream (subscript 1) and downstream (subscript 2) properties:

• Mass: $\rho_1 u_1 = \rho_2 u_2$
• Momentum: $p_1 + \rho_1 u_1^2 = p_2 + \rho_2 u_2^2$
• Energy: $h_1 + \frac{1}{2}u_1^2 = h_2 + \frac{1}{2}u_2^2$

These jump conditions serve as internal boundary conditions for numerical methods that need to capture or fit shock waves.

Oblique vs normal shocks

• Normal shocks are perpendicular to the flow. The entire velocity change is in the flow direction, and the downstream flow is always subsonic.
• Oblique shocks are inclined at an angle to the flow. They change both the magnitude and direction of the velocity, and the downstream flow can remain supersonic.
• The oblique shock relations include an additional geometric parameter (the wave angle), and different boundary condition treatments apply depending on whether the shock is attached or detached from the body.

Shock-boundary layer interaction

When a shock wave impinges on a boundary layer, the sudden adverse pressure gradient can cause the boundary layer to thicken or separate entirely. This interaction produces:

• Flow separation and recirculation zones
• Unsteady oscillations (buffet on transonic airfoils)
• Locally increased heat transfer rates

Accurate simulation requires resolving both the shock and the boundary layer simultaneously, which demands fine mesh resolution near the wall and robust boundary condition treatment for both phenomena.

Computational fluid dynamics (CFD)

CFD solves the governing equations of fluid motion numerically, and boundary conditions are one of the most important inputs you provide. A simulation with incorrect or inconsistent boundary conditions will produce wrong answers regardless of how fine your mesh is or how advanced your solver is.

Mesh generation techniques

The mesh must represent the geometry accurately enough that boundary conditions can be applied correctly:

• Boundary-fitted meshes conform to the geometry, so boundary faces align with physical surfaces. This makes applying wall conditions straightforward.
• Cartesian cut-cell methods use a simple background grid and cut cells at the geometry boundary. They handle complex geometries automatically but require special treatment at cut cells.
• Overset (Chimera) grids use multiple overlapping meshes, each with its own boundary conditions. Interpolation between grids handles the communication. This simplifies meshing for moving bodies or complex multi-component geometries.

Boundary condition specification in CFD

In practice, you specify boundary conditions for each variable (velocity, pressure, temperature, turbulence quantities) on every boundary face of the domain. Steps for a typical external aerodynamics simulation:

1. Identify each boundary type: wall, inlet, outlet, far-field, symmetry.
2. Assign the appropriate condition type and values for each variable on each boundary.
3. Check compatibility: for example, you generally shouldn't fix pressure at both the inlet and outlet for an incompressible solver.
4. Use user-defined functions (UDFs) for non-standard conditions like time-varying inflow profiles or coupled thermal boundaries.

Wall functions and turbulence modeling

Fully resolving the viscous sublayer of a turbulent boundary layer requires extremely fine mesh spacing near the wall ($y^+ \approx 1$). Wall functions provide an alternative by using semi-empirical formulas to bridge the gap between the wall and the first mesh point in the log-law region ($y^+ \approx 30\text{–}300$).

Common turbulence models and their wall boundary condition requirements:

• $k\text{-}\varepsilon$ model: Typically used with wall functions. Requires boundary values for $k$ and $\varepsilon$ at the wall.
• $k\text{-}\omega$ model: Can integrate directly to the wall without wall functions. Requires $\omega$ to be specified at the wall (often as a large value).
• SST (Shear Stress Transport) model: Blends $k\text{-}\varepsilon$ in the freestream with $k\text{-}\omega$ near the wall. Handles both fine and coarse near-wall meshes.

Choosing the right wall treatment directly affects predictions of separation, skin friction, heat transfer, and drag.

Applications in aerodynamics

Airfoil and wing design

• No-slip and no-penetration on the airfoil surface capture viscous drag and pressure distribution.
• The Kutta condition at the trailing edge enforces smooth flow departure, which determines the circulation and therefore the lift.
• Freestream conditions at the far-field set the Mach number, angle of attack, and thermodynamic state.

Wind tunnel testing

• Inflow conditions match the wind tunnel test section: velocity magnitude, turbulence intensity, and length scale.
• Outflow conditions are set to minimize blockage effects and prevent artificial upstream influence.
• Wall conditions on the tunnel walls can be slip (to simulate an unbounded flow) or no-slip (to capture wall interference), depending on the study goals.

Aircraft performance analysis

• Full aircraft configurations require boundary conditions for engine inlets (mass flow or total pressure) and exhausts (total temperature, mass flow).
• Actuator disk or blade element models represent propellers and rotors as momentum and energy sources rather than resolving every blade, simplifying the boundary condition treatment.
• Atmospheric conditions (pressure, temperature, density) vary with altitude and must be set appropriately for each flight condition being analyzed.