Boundary layer — Prandtl's concept, growth along flat plate: Visual walkthrough

Step 1 — The plate, the stream, and the two speeds

WHAT. Draw a flat plate lying flat, with fluid sweeping over it left-to-right at one steady speed. Call that far-away speed $U$ (units: metres per second, m/s). Now zoom into the fluid touching the plate.

WHY. Before any maths, we must fix the single physical fact that makes a boundary layer exist at all: the fluid glued to the wall cannot slide. This is the no-slip condition — the layer of fluid molecules right against the solid has the same velocity as the solid, which for a stationary plate is zero.

PICTURE. In the figure, the amber arrows far above the plate are all the same length ( $= U$ ). The arrow right on the plate has zero length. So we already have two speeds living very close together: $0$ at the wall, $U$ up high.

Figure — Boundary layer — Prandtl's concept, growth along flat plate

Here $y$ means height measured straight up from the plate surface; $y = 0$ is the wall itself.

Step 2 — The velocity profile: a squeeze from 0 to $U$

WHAT. Plot $u$ against height $y$ at one fixed downstream position. It starts at $0$ on the wall and bends up to $U$ .

WHY. We need to see where the change from $0$ to $U$ happens. It does not happen everywhere — it happens across a thin band. The vertical size of that band is the whole quantity we are chasing.

PICTURE. The cyan curve rises steeply near the wall, then flattens as it nears $U$ . Mark the height where it has essentially reached $U$ — that height is $δ$ .

The steepness of that cyan curve near the wall is a rate of change: how much speed $u$ gains per unit rise $y$ . We write it $\frac{\partial u}{\partial y}$ .

Step 3 — Why viscosity refuses to be ignored here

WHAT. Attach a physical force to that steep gradient. Neighbouring fluid layers moving at different speeds rub on each other; the rubbing stress is $τ = μ \frac{\partial u}{\partial y}$ (Newton's law of viscosity).

WHY — the tool choice. Why bring in $μ$ at all when air's $μ$ is microscopically small? Because the stress is $μ$ times the gradient, not $μ$ alone. In Step 2 we saw the gradient near the wall is $\sim U / δ$ , and $δ$ is tiny, so the gradient is enormous. A microscopic $μ$ multiplied by an enormous gradient gives a stress that is not negligible. That is the resolution of d'Alembert's paradox — the drag lives in this thin band, not in the outer flow.

PICTURE. Two sliding sheets of fluid; the amber shear arrows point opposite ways; the label shows small $μ$ but a fat gradient producing a real force.

Step 4 — Repackage viscosity as diffusion: meet $ν$

WHAT. Divide viscosity $μ$ by the fluid's density $ρ$ (mass per volume, kg/m³). Call the result the kinematic viscosity $ν = μ / ρ$ .

WHY — the tool choice. We are about to argue with a diffusion picture (Step 5), and diffusion needs a coefficient measured in area-per-time (m²/s), not in Pa·s. Check the units: $ν = \frac{μ}{ρ} = \frac{Pa\cdots}{kg/m ^{3}} = \frac{kg / ( m\cdots )}{kg / m ^{3}} = \frac{m ^{2}}{s}$ It comes out exactly as "area per time" — the signature of a spreading process. So $ν$ is the natural currency for "how fast does the wall's slowing-down spread upward."

PICTURE. A units-cancellation ledger: $μ$ (thick), $ρ$ (a stack of mass in a box), and the resulting $ν$ tagged m²/s, next to a small dye blob spreading — the visual meaning of a diffusion coefficient.

Step 5 — How far can slowing-down spread in the time available?

WHAT. Two questions, combined. (a) A fluid parcel travels the plate length $x$ at speed $\sim U$ , so it is exposed to the wall for a time $t \sim x / U$ . (b) During that time, momentum-diffusion (coefficient $ν$ ) spreads the wall's influence a height $δ \sim ν t$ .

WHY — the tool choice. Why $ν t$ and not $ν t$ ? Because any diffusion spreads as the square root of time: a dye blob of radius growing like $D t$ , heat creeping like $α t$ , and here momentum like $ν t$ . Diffusion is a random-walk spreading — doubling the spread costs four times the time. That single fact is the whole reason $δ$ ends up $\propto x$ and never linear.

PICTURE. Top: a parcel riding the stream, a clock showing $t = x / U$ . Bottom: the shaded slowed-region fanning upward as $ν t$ , its ceiling being $δ$ .

Step 6 — Combine into the growth law

WHAT. Substitute $t = x / U$ into $δ \sim ν t$ .

WHY. This is the whole point: the exposure time depends on position $x$ , so the spread depends on $x$ . Plug and simplify.

$δ \sim ν \cdot \frac{x}{U} = \frac{ν x}{U}$

PICTURE. The finished sideways-parabola: $δ$ starting at $0$ at the leading edge and curving up as $x$ — never a straight ramp.

The exact laminar (laminar) constant from the Blasius solution of the Navier–Stokes equations sharpens $\sim$ to $\approx$ : $δ (x) \approx \frac{5.0 x}{R e _{x}}$

Step 7 — Edge cases: does the picture survive the extremes?

WHAT. Test the formula where it might break: the leading edge, a still fluid, an inviscid fluid, and a very fast stream.

WHY. A law you trust must give sensible pictures at every corner, including degenerate ones — otherwise the reader hits a wall the derivation never showed.

PICTURE. Four mini-panels, each an extreme, each matched to the formula's prediction.

Corner case	Formula says	Picture / meaning
Leading edge, $x = 0$	$δ = ν \cdot 0/ U = 0$	The layer is born with zero thickness — fluid has had no time to be slowed.
Inviscid fluid, $ν \to 0$	$δ \to 0$	No stickiness, no slowing spreads — no boundary layer. Exactly [[d'Alembert's paradox
Very fast stream, $U \to \infty$	$δ \to 0$	Parcels blow past before diffusion can climb — vanishingly thin layer.
Slow / still stream, $U \to 0$	$δ \to \infty$	Huge exposure time; the "thin layer" idea breaks down — Prandtl's split needs high $R e_{x}$ .

Step 8 — Bonus picture: where friction is fiercest

WHAT. Feed $δ$ back into the wall stress $τ_{w} = μ \partial u / \partial y ∣_{y = 0} \sim μU / δ$ .

WHY. The same picture that gave $δ$ also tells us the skin-friction along the plate — for free.

$τ_{w} \sim μ \frac{U}{δ} \sim μ \frac{U}{ν x / U} = μU \frac{U}{ν x} \propto \frac{1}{x}$

PICTURE. The wall-stress curve dropping as $1/ x$ : steepest velocity slope (thinnest $δ$ ) at the leading edge means friction is largest there and fades downstream.

The one-picture summary

Everything on one canvas: the plate, the growing $x$ layer, the velocity profile squeezing $0 \to U$ , the diffusion arrow climbing as $ν t$ , and the exposure-time clock — the four ideas that chain into $δ \sim ν x / U$ .

Recall Feynman retelling of the whole walkthrough

Picture dragging your hand along a table covered in honey. The honey touching the table is stuck dead-still — that's no-slip (Step 1). A little higher it moves a bit, higher still it flows freely at speed $U$ ; that squeeze from stuck to free happens across a thin band whose height is $δ$ (Step 2). Because the whole speed jump is packed into a tiny height, the honey layers are shearing hard against each other, and that rubbing force is real even for slightly-sticky fluid — small stickiness times a huge slope (Step 3). Now think of "stuckness" as something that spreads upward like a dye blob, at a rate set by $ν$ measured in area-per-time (Step 4). A honey parcel is only over any patch of table for a short time $t = x / U$ , and in that time the stuckness climbs a height $ν t$ (Step 5). Put those together — $δ \sim ν x / U$ — and you get a slow-zone that fattens downstream but slower and slower, like a stretched sideways parabola (Step 6). At the very start it's zero-thick; kill the stickiness or crank the speed and it vanishes (Step 7); and the friction bites hardest right at the front where the layer is thinnest (Step 8). That's the entire boundary layer, built from stuck honey.