Dimensional Analysis and Introduction to Boundary Layers

TL;DR

You'll learn how to strip units out of the Navier-Stokes equations to reveal the Reynolds number, the single parameter that controls whether a flow behaves like it has no friction or like it's dominated by it. You'll then see why, even at huge Reynolds number, friction never fully disappears — it hides in a thin layer near walls. By the end you can estimate that layer's thickness with one formula.

1. The Mental Model

Every fluid flow is a tug-of-war between inertia (stuff wants to keep moving) and viscosity (stuff wants to stop moving relative to its neighbors). The Reynolds number is just the score of that tug-of-war. When inertia wins almost everywhere, viscosity still wins in a thin strip next to any solid surface, because that's the only place velocity gradients get steep enough to matter. The boundary layer is where the loser of the global tug-of-war quietly wins a local rematch.

2. The Core Material

Nondimensionalizing the Navier-Stokes Equations

Photo by Monstera Production on Pexels

Start with the incompressible Navier-Stokes equations:

$$\frac{\partial \mathbf{u}}{\partial t} + (\mathbf{u}\cdot\nabla)\mathbf{u} = -\frac{1}{\rho}\nabla p + \nu \nabla^2 \mathbf{u}$$

Every term here has dimensions. The trick of dimensional analysis is to pick one characteristic length $L$, one characteristic velocity $U$, and build every other scale from them: a time scale $T = L/U$, and a pressure scale $\rho U^2$ (this comes from Bernoulli-type reasoning — pressure differences that drive flow scale as density times velocity squared).

Define dimensionless variables:

$$\mathbf{x}^* = \frac{\mathbf{x}}{L}, \quad \mathbf{u}^* = \frac{\mathbf{u}}{U}, \quad t^* = \frac{tU}{L}, \quad p^* = \frac{p}{\rho U^2}$$

Substitute these into the equation. Every derivative picks up a scale factor: $\nabla = \frac{1}{L}\nabla^*$, $\partial_t = \frac{U}{L}\partial_{t^*}$. After the dust settles (and you divide everything by $U^2/L$, the common factor on the inertia terms), you get:

$$\frac{\partial \mathbf{u}^*}{\partial t^*} + (\mathbf{u}^*\cdot\nabla^*)\mathbf{u}^* = -\nabla^* p^* + \frac{1}{Re}\nabla^{*2}\mathbf{u}^*$$

where

$$Re = \frac{UL}{\nu}$$

That's the entire payoff of the exercise: one dimensionless number, the Reynolds number, controls the relative weight of viscosity in the equation. Every other detail of the geometry and fluid — density, viscosity, speed, size — collapses into this single ratio. Two totally different flows (water around a submarine, air around a wing) behave identically in dimensionless form if they share the same $Re$ and geometry. This is why wind-tunnel testing with scaled models works at all.

The Reynolds Number and Its Physical Meaning

Photo by Markus Spiske on Pexels

$Re = UL/\nu$ is literally the ratio of the inertia term to the viscous term, evaluated at the outer length scale $L$:

$$Re \sim \frac{|(\mathbf{u}\cdot\nabla)\mathbf{u}|}{|\nu \nabla^2 \mathbf{u}|} \sim \frac{U^2/L}{\nu U/L^2} = \frac{UL}{\nu}$$

Two limits matter enormously:

• $Re \to 0$ (Stokes flow): viscosity dominates everywhere. The $1/Re$ term blows up, so you drop the inertia terms instead and get the linear Stokes equations. Think of a bacterium swimming in honey.
• $Re \to \infty$ (Euler flow): the $1/Re$ term vanishes, and you're left with the inviscid Euler equations, $\partial_t \mathbf{u} + (\mathbf{u}\cdot\nabla)\mathbf{u} = -\nabla p/\rho$. This looks great — no viscosity, no messy diffusion — except for one glaring problem: Euler's equations can't enforce the no-slip condition ($\mathbf{u}=0$) at a solid wall. Inviscid flow is happy to slide along a surface at full speed. Real fluids don't do that.