What is a boundary layer? Why is it important in CFD?

The boundary layer is a thin region near a wall where flow velocity changes rapidly from zero (at the wall) to the free-stream velocity. Most of the drag and heat transfer occur in this region, making accurate resolution critical for CFD accuracy.

What is the difference between laminar and turbulent boundary layers?

Laminar boundary layers have smooth parabolic velocity profiles and grow slowly. Turbulent boundary layers have a very thin viscous sublayer near the wall and rapid growth due to turbulent mixing. Turbulent boundary layers produce much higher wall shear stress and heat transfer coefficients.

What is y+ and what values should I use?

y+ is a dimensionless wall distance indicating where the first mesh cell center is located relative to the wall region. For wall functions, y+ ≈ 30–300; for wall-resolved approaches, y+ ≈ 1. The requirement depends on the turbulence model used.

What is the difference between wall functions and wall-resolved approaches?

Wall functions use empirical correlations to bridge the viscous sublayer, reducing mesh requirements but losing accuracy in separated flows. Wall-resolved approaches mesh the entire sublayer for higher accuracy, requiring many more cells. Wall-resolved is recommended for separation and adverse pressure gradients.

How should I set up boundary layer inflation layers?

First, calculate the first layer height from your target y+ value. Use a growth rate of 1.1–1.2 and 15–25 layers to extend beyond the boundary layer thickness. Monitor aspect ratios and transition smoothly to the bulk mesh.

Boundary Layer — CAE Glossary

Category: Glossary | 2026-03-28

CAE visualization for boundary layer - technical simulation diagram

What is a Boundary Layer?

🧑‍🎓

I keep hearing about "boundary layer" in CFD discussions. What exactly is it? Something to do with the wall, right?

🎓

In short, it's the thin region right next to a wall where fluid velocity changes rapidly from zero (stuck to the wall) to the free-stream velocity. Inside this layer, viscous effects are huge, and that's where most of your drag and heat transfer come from. For an airplane wing, for example, the skin friction drag comes entirely from the velocity gradient inside the boundary layer.

🧑‍🎓

Why is velocity zero at the wall? That seems odd.

🎓

That's the no-slip condition, a fundamental principle in viscous fluid mechanics. Fluid molecules stick to the wall, and then they drag the adjacent layer along, creating a velocity gradient. The distance over which this velocity recovery happens is what we call the boundary layer. Prandtl introduced this concept back in 1904, and it's been central to fluid mechanics ever since.

The boundary layer thickness $\delta$ is conventionally defined as the wall distance where velocity reaches 99% of the free-stream value $U_\infty$. For laminar flow over a flat plate, the Blasius solution gives:

$$\delta \approx \frac{5.0\,x}{\sqrt{Re_x}}, \qquad Re_x = \frac{U_\infty\,x}{\nu}$$

where $x$ is distance from the leading edge and $\nu$ is kinematic viscosity.

Laminar vs. Turbulent Boundary Layers

🧑‍🎓

Boundary layers can be laminar or turbulent? What's the difference?

🎓

Big difference. In laminar boundary layers, fluid flows in smooth, orderly layers with velocity profiles that look like smooth parabolas. The layer grows slowly. In turbulent ones, except for a paper-thin viscous sublayer right at the wall, everything is churned by turbulent eddies. The velocity profile is much fuller — it rises steeply near the wall then levels off. Wall friction and heat transfer in turbulent layers are much, much higher.

🧑‍🎓

So turbulent is always worse then? More drag? How common is it in real problems?

🎓

Yes, higher friction. But turbulence is the norm in industrial flows because Reynolds number is usually very high. A car at highway speed? Re ≈ 10⁶–10⁷. Almost the entire surface is turbulent except right at the leading edge. Microfluidics and MEMS? Those might stay laminar because Re ≈ 1–100. But for most aerodynamics and hydrodynamics, you're dealing with turbulent boundary layers.

The turbulent boundary layer thickness follows an empirical power law (1/7-law based):

$$\delta \approx \frac{0.37\,x}{Re_x^{1/5}}$$

The internal structure of a turbulent boundary layer divides into three regions based on dimensionless wall distance $y^+$:

Viscous sublayer: $y^+ < 5$ — Viscous stress dominates. Velocity follows $u^+ = y^+$ (linear).
Buffer layer: $5 < y^+ < 30$ — Transition region where viscous and turbulent stresses balance.
Log-law region: $30 < y^+ < 300\text{–}500$ — Turbulent stress dominates. Velocity follows the log law.

$$u^+ = \frac{1}{\kappa}\ln y^+ + B \qquad (\kappa \approx 0.41,\; B \approx 5.2)$$

y⁺ (y-plus) Fundamentals

🧑‍🎓

I see y⁺ all the time in CFD setup. What exactly is it?

🎓

It's a dimensionless wall distance defined as:

$$y^+ = \frac{y\,u_\tau}{\nu}, \qquad u_\tau = \sqrt{\frac{\tau_w}{\rho}}$$

Here $y$ is distance from the wall, $u_\tau$ is friction velocity, $\tau_w$ is wall shear stress, $\rho$ is density, and $\nu$ is kinematic viscosity. In CFD, we care about where the center of your first mesh cell sits on this scale. That directly determines whether your turbulence model will work correctly and how much mesh you need near the wall.

🧑‍🎓

What y⁺ should I actually use? I've heard "always use y⁺=1" but I'm not sure when.

🎓

That depends entirely on your turbulence model:

With wall functions (Standard k-ε and similar): $y^+ = 30\text{–}300$. You skip the viscous sublayer entirely with empirical formulas, so if $y^+$ gets too small, those formulas break down.
Wall-resolved (SST k-ω, Spalart-Allmaras, etc.): $y^+ \approx 1$. You mesh all the way into the viscous sublayer.

The most popular industrial choice, SST k-ω, recommends $y^+ \approx 1$. Many solvers auto-switch between wall-function and wall-resolved modes, so anywhere from $y^+ = 1$ to $y^+ = 5$ usually works. But check your solver's documentation first.

Wall Functions vs. Wall-Resolved Approaches

🧑‍🎓

Wall functions let you use coarser mesh near walls, right? That saves computation?

🎓

Half right, half dangerous thinking. Wall functions use the log-law to skip the viscous sublayer, so you can place your first cell in the log-law region ($y^+ = 30\text{–}300$) instead of inside the sublayer. That cuts computational cost significantly. But "coarse" only applies to the wall-normal direction. Along the wall (streamwise and spanwise), you still need fine resolution to capture flow features. So it's not a free lunch.

🧑‍🎓

When can't you use wall functions?

🎓

Wall functions assume the log-law holds. That breaks down in several important cases:

Flow separation and reattachment: Adverse pressure gradients destroy the log-law assumptions. Car rear windows and airfoil stall are classic examples.
Rapid acceleration or deceleration: Nozzles and diffusers with strong streamwise pressure changes.
Accurate heat transfer prediction: Electronics cooling, turbine blades — when wall temperature gradients matter.
Transition modeling: If you're capturing laminar-to-turbulent transition.

In those cases, bite the bullet and use $y^+ \approx 1$ wall-resolved mesh. Yes, more elements. But wrong answers from wall functions are much worse.

Boundary Layer Mesh in Practice

🧑‍🎓

How do I actually set up inflation layers? First cell height calculation stumps me every time.

🎓

Start by calculating first layer height $\Delta s$ from your target y⁺. You need wall shear stress first. A practical approach uses empirical friction coefficient:

$$C_f \approx \frac{0.058}{Re_L^{0.2}}$$ $$\tau_w = \frac{1}{2}\,C_f\,\rho\,U_\infty^2$$ $$\Delta s = \frac{y^+\,\mu}{\sqrt{\rho\,\tau_w}}$$

For air ($\rho=1.225$ kg/m³, $\mu=1.8\times10^{-5}$ Pa·s) at $U_\infty=30$ m/s with $L=1$ m and $y^+=1$, you get roughly $\Delta s \approx 0.015$ mm.

🧑‍🎓

What about growth rate and number of layers? I usually just guess 1.2 and 10 layers…

🎓

Many people do guess (laugh), but there are actual guidelines. Growth rate should be 1.1–1.2. Above 1.3, cell quality collapses fast and you get numerical diffusion and convergence problems. Layer count should be enough to reach or exceed boundary layer thickness $\delta$, so typically 15–25 layers. Some meshers call it "number of divisions" instead, but the idea is the same.

🧑‍🎓

After I run the analysis, how do I check if my y⁺ is actually correct?

🎓

Nearly every CFD solver can output wall $y^+$ in post-processing. OpenFOAM has the yPlus utility. Fluent lets you plot Wall Y+ as a contour. What matters is that y⁺ across the entire wall surface is in your target range. If some spot is way off, refine mesh there. Before trusting any results, always verify y⁺ distribution — that's the golden rule.

Reynolds Number — Ratio of inertial to viscous forces. Determines boundary layer development and transition.
y⁺ — Dimensionless wall distance. The most critical parameter for boundary layer mesh design.
Wall Function — Empirical model that approximates the viscous sublayer.
Turbulence Model — Set of equations that close the RANS equations. Directly tied to boundary layer handling.
Inflation Layer — Thin layered mesh perpendicular to wall surfaces for boundary layer resolution.
Law of the Wall — Universal logarithmic velocity profile in the outer boundary layer region.
Separation — Boundary layer departure from the wall due to adverse pressure gradient.

Precise understanding of CAE terminology is the foundation for team communication. — Project NovaSolver supports learning for practical engineers.

Struggling with boundary layer mesh setup?

Project NovaSolver focuses on solving real challenges CAE engineers face daily — mesh refinement iteration, y⁺ tuning, wall function trade-offs. Your practical experience directly shapes better tool development.