Thermal Boundary Layer
Thermal Boundary Layer: Theoretical Foundations
Fundamental Concepts of Thermal Boundary Layer
Professor, how is the thermal boundary layer related to the velocity boundary layer?
When fluid flows over a heated (or cooled) wall surface, a thin layer where the temperature changes rapidly forms near the wall. This is the thermal boundary layer. Similar to the velocity boundary layer, it asymptotically approaches the freestream temperature as you move away from the wall. The ratio of their thicknesses is determined by the Prandtl number $Pr$.
If $Pr > 1$ (water, oil, etc.), the thermal boundary layer is thinner than the velocity boundary layer. If $Pr < 1$ (air, molten metal, etc.), the thermal boundary layer is thicker.
What is the $Pr$ of air?
For air at room temperature, $Pr \approx 0.71$, so the thermal boundary layer is slightly thicker than the velocity boundary layer. For water, $Pr \approx 7$, so the thermal boundary layer becomes about half the thickness of the velocity boundary layer. For engine oil, $Pr \sim 1000$ or more, making the thermal boundary layer extremely thin.
Theoretical Solution for Laminar Thermal Boundary Layer
Are there analytical solutions?
For laminar boundary layers on an isothermal flat plate, the Blasius solution (velocity field) and the Pohlhausen solution (temperature field) are classical theoretical solutions. The local Nusselt number is given by
(for $Pr > 0.6$). This formula is always used in basic CFD validation. You should confirm that the wall Nusselt number from CFD matches this theoretical solution within 2-3% before proceeding to more complex problems.
What about turbulent flow?
For turbulent boundary layers, $Nu_x = 0.0296 Re_x^{4/5} Pr^{1/3}$ is an empirical correlation. In turbulence, the transport of momentum and heat is dominated by eddy diffusion (turbulent diffusivity), so the thickness ratio between the thermal and velocity boundary layers is determined by the turbulent Prandtl number $Pr_t \approx 0.85$ to $0.9$.
The Foundation of Thermal Boundary Layer Theory—Pohlhausen's Integral Method (1921)
Seventeen years after Prandtl (1904) proposed the concept of the velocity boundary layer, his student E. Pohlhausen (1921) derived the "integral equation for the thermal boundary layer," providing the first analytical solution for the temperature distribution in forced convection over a flat plate. His analysis revealed the dependence Nu ∝ Re^(1/2)×Pr^(1/3) and showed that the boundary layer thickness ratio (δ_t/δ) differs by three orders of magnitude between low-Pr liquid metals (Pr≈0.01) and high-Pr oils (Pr≈1000). This simple scaling law is still used today, a century later, to check the validity of CFD results. In practice, it's common sense to regard CFD output that deviates from this analytical solution as a sign of problems with the mesh or material properties.
Computational Methods for Thermal Boundary Layer
Wall Mesh Requirements
How should I design the mesh to accurately resolve the thermal boundary layer?
Wall-normal mesh design is closely related to the wall treatment of the turbulence model. For the Low-Reynolds number approach (direct resolution), a guideline is to place the first cell at $y^+ \approx 1$ and have at least 5 layers within the viscous sublayer and 5-10 layers in the buffer layer.
Do the thermal and velocity boundary layers require different mesh densities?
For fluids with $Pr > 1$, the thermal boundary layer is thinner than the velocity boundary layer, so accurately predicting heat transfer may require an even finer mesh than for the velocity field. Specifically, $y^+_{T} = y u_\tau / \alpha < 1$ should be satisfied, which corresponds to $y^+ \cdot Pr < 1$. For $Pr = 7$ (water), $y^+ < 0.14$ is ideal.
Is such a thin first cell practically achievable?
In practice, sufficient accuracy is often obtained with $y^+ \approx 0.5$ to $1$. Even for water, if $y^+ < 1$ is satisfied, the error in the Nusselt number can be kept within 5%. What's important is the growth ratio in the wall-normal direction, which should be kept within 1.1 to 1.2.
How to Check y+ Values
How do I check if $y^+$ is appropriate after the calculation?
What if $y^+$ is too large?
Reduce the thickness of the first layer of the inflation layer (prism layer) or increase the number of layers. However, when flow velocity varies (e.g., different velocities near the inlet and downstream), it can be difficult to satisfy $y^+ < 1$ over the entire wall. Fluent's Enhanced Wall Treatment and STAR-CCM+'s All y+ Wall Treatment automatically switch wall treatments based on $y^+$ values, so using these is safer in practice.
Handling High-Pr Fluids
What about high-Pr fluids like oil ($Pr > 100$)?
Standard wall functions significantly lose accuracy as $Pr$ increases. Fluent's Enhanced Thermal Wall Treatment includes a high-Pr correction, so this should be enabled. In OpenFOAM, alphatJayatillekeWallFunction contains Jayatilleke's (1969) correction. For the mesh, the golden rule is to make $y^+$ as small as possible.
Thermal Boundary Layer y+ Management—The Branching Point Between Wall Functions and Low-Re Solutions is y+=1
The accuracy of the thermal boundary layer is first determined by the first cell height (y+ value). The wall function approach (y+=30-300) assumes the logarithmic law region to estimate wall heat flux, so errors increase rapidly in regions with strong pressure gradients or separation. On the other hand, when using low-Reynolds number models (SST-ω, v2-f, etc.), y+<1 is required, and 15-20 prism layer cells must be secured to resolve the steep temperature gradient in the viscous sublayer. In practice, a staged refinement strategy—"first check the global flow with wall functions → switch to low-Re only for important heat flux surfaces"—is the standard way to balance computational cost and accuracy.
Thermal Boundary Layer in Practice
Heat Transfer in Transitional Boundary Layers
How does heat transfer change when transition from laminar to turbulent flow occurs?
The Nusselt number increases sharply at the transition point. This is because turbulent mixing thins the thermal boundary layer and steepens the temperature gradient at the wall. For aircraft wing leading edges or wind turbine blades, the transition location determines the external surface temperature distribution, so accurate prediction is crucial.
How can I predict transition in CFD?
The Transition SST ($\gamma$-$Re_\theta$) model is the standard choice. The freestream turbulence intensity $Tu$ at the inlet greatly affects the transition location, so it must be set accurately to match experimental conditions. For turbine airfoil CFD, sensitivity analysis within the range $Tu = 1$ to $10$% is common practice.
Heat Transfer with Separation
What happens to the thermal boundary layer in separated flows like over a backward-facing step?
Immediately after the separation point, the flow velocity near the wall decreases, reducing the Nusselt number. At the reattachment point, a flow structure similar to jet impingement forms, and the Nusselt number peaks. Downstream of the reattachment point, as the boundary layer redevelops, the Nusselt number gradually approaches the fully developed value.
In such cases, the thermal boundary layer becomes highly sensitive to small mesh or turbulence model variations, so careful validation against experimental data is essential.