Boundary layer separation — adverse pressure gradient: Visual walkthrough

Step 1 — What is a velocity profile? (the picture everything rests on)

The crucial fact: $u$ is not one number. It changes with $y$ . Right at the wall the fluid is stuck (glued by friction); far out it moves at the full free-stream speed $U$ . A graph of $u$ against $y$ is called a velocity profile.

Figure s01 below is the anchor for the whole page: it is the picture we will keep re-drawing as the flow marches downstream.

Figure — Boundary layer separation — adverse pressure gradient

Read s01 like this: the horizontal axis is speed $u$ , the vertical axis is height $y$ above the wall. Follow the yellow curve — it leans over near the ground (slow, short blue arrows) and stands up tall higher up (fast, long blue arrows), until it reaches the free stream $U$ . The pink dot pinned to the origin is the no-slip point. That leaning shape is the whole story.

Step 2 — The slope of the profile is the wall friction

The slope of $u$ as you move up in $y$ is written $\frac{\partial u}{\partial y}$ .

Now why do we care about the slope right at the wall? Because friction is a shearing thing — one layer of fluid dragging on the layer beside it. The harder the speed changes across a gap, the harder the layers rub. That rubbing force per unit area is the wall shear stress:

In figure s02 the pink dashed line is the tangent to the profile at the wall — its steepness is exactly $\partial u / \partial y ∣_{0}$ . Watch that dashed line as we go: the pedagogical target of the whole page is the moment this dashed tangent lies flat (slope zero).

Why the tool $\partial u / \partial y$ and not just "speed"? Because friction does not care how fast the flow is far away — it cares how fast speed changes across the tiny near-wall gap. That change is a slope, so a slope (derivative) is exactly the right tool. This little formula is the hero of the page: when this slope reaches zero, the flow lets go of the wall. We spend the rest of the page proving that.

Step 3 — The momentum equation inside the thin layer

We need a law that governs $u$ everywhere in the layer, not just at the wall. That law is Newton's second law ("mass × acceleration = force") written for a fluid particle inside the boundary layer.

In s03 the yellow equation is colour-tagged term-by-term: blue = acceleration (what the fluid is doing), pink = pressure push, white = viscous smoothing. This is just $F = ma$ dressed up in fluid clothes.

Step 4 — Evaluate the law on the wall itself

Here is the clever move. Set $y = 0$ — sit exactly on the surface — and watch two terms die. But note carefully: two different boundary conditions are doing the killing, not one.

Now look at the left side of the momentum equation with both conditions applied:

$u \frac{\partial u}{\partial x} + v \frac{\partial u}{\partial y} ⟶ u = 0 (no-slip) 0 \cdot \frac{\partial u}{\partial x} + v = 0 (impermeability) 0 \cdot \frac{\partial u}{\partial y} = 0$

The first term is multiplied by $u$ , which is $0$ by no-slip → gone.
The second term is multiplied by $v$ , which is $0$ by impermeability → gone.

So at the wall the acceleration terms vanish entirely. Newton's law collapses to a clean balance between the two right-hand forces:

$- \frac{1}{ρ} \frac{d p}{d x} + ν \frac{\partial ^{2} u}{\partial y ^{2}}_{y = 0} = 0$

In figure s04 the two vanishing terms are struck through and labelled with the condition that kills each one, so you can see why the left side collapses.

Why this step is the whole trick: everywhere else in the layer the messy convective terms fight us. Exactly on the wall, the two wall conditions erase them for free, leaving only pressure vs. viscosity. That is the cleanest possible place to stand.

Step 5 — The wall-curvature law

Now tidy up. Multiply the balance through by $ρ$ and use $μ = ρ ν$ :

$- \frac{d p}{d x} + ρ ν \frac{\partial ^{2} u}{\partial y ^{2}}_{y = 0} = 0 ⟹ μ \frac{\partial ^{2} u}{\partial y ^{2}}_{y = 0} = \frac{d p}{d x}$

This boxed result says something beautiful: the pressure gradient single-handedly decides how the velocity profile bends right at the wall. No guessing — the sign of $d p / d x$ is the sign of the wall curvature.

Figure s05 frames the boxed law and shows the two cup/cap shapes it selects — refer back to it whenever the sign of $d p / d x$ comes up.

Why we care about curvature and not just slope: the slope tells us the friction now; the curvature tells us whether the profile is about to fill out or hollow out — the early warning of trouble.

Step 6 — Favourable gradient: the profile stays fat and healthy

Take $\frac{d p}{d x} < 0$ — pressure falling along the wall (a favourable gradient; see Diffusers and nozzles for the opposite duct).

From the boxed law, negative $d p / d x$ means the wall curvature $\partial^{2} u / \partial y^{2} ∣_{0}$ is negative: the profile caps over (⌢) from the very bottom. So the curve rushes up steeply from the wall and rounds smoothly into the free stream $U$ — no bending back, no inflection.

Steep at the wall → large $\partial u / \partial y ∣_{0}$ → large $τ_{w}$ → strong forward grip.
The near-wall fluid is being helped along; it will not stop.

In figure s06, notice the blue dashed tangent at the wall is steep, and the yellow curve is fat all the way up — compare it side-by-side with the sick profile coming in s07.

This is a happy, attached boundary layer. Nothing separates.

Step 7 — Adverse gradient: an inflection point is forced into the profile

Now the villain: $\frac{d p}{d x} > 0$ — pressure rising (the adverse gradient).

The boxed law now makes the wall curvature positive: near the wall the profile bends away (⌣). We must now argue there is a region of negative curvature higher up — here is the honest chain:

At the wall, curvature is positive: $\frac{\partial ^{2} u}{\partial y ^{2}}_{0} = \frac{1}{μ} \frac{d p}{d x} > 0$ .
Far out, the flow settles onto the constant free stream, $u \to U$ . A flat line has zero slope and zero curvature, so $\frac{\partial ^{2} u}{\partial y ^{2}} \to 0$ as $y$ grows large.
Approaching that flat top, the slope $\partial u / \partial y$ is positive (still climbing) but must decrease to zero. A positive quantity that is decreasing has a negative rate of change — and the rate of change of the slope is the curvature. So just below the free stream the curvature is negative.
Curvature is therefore positive at the wall and negative just under the free stream. A continuous quantity that changes sign must pass through zero in between — an inflection point.

An inflection point is a known danger sign: the S-shaped profile carries a pocket of very slow, easily-reversed fluid near the wall. The adverse pressure — a force pushing backward on that already-slow fluid — is what tips it over.

Figure s07 marks all three regions explicitly: the cup (⌣) below, the blue inflection dot where curvature $= 0$ , and the cap (⌢) above where the curve rounds into $U$ .

Why this connects to the fuel-tank picture: the energy-poor wall fluid (slow, from viscosity) meets a backward push (rising pressure). The inflection point is the mathematical fingerprint of that mismatch. See Flow over a cylinder and sphere for where this bites on a real body, and Stall on an aerofoil for the wing version.

Step 8 — The separation point: wall slope reaches zero

Follow the flow downstream under a steadily adverse gradient. The near-wall slope $\partial u / \partial y ∣_{0}$ shrinks and shrinks (the profile leans more and more upright at the bottom). The exact instant it reaches zero is separation:

Just downstream of this point the slope goes negative: $\frac{\partial u}{\partial y}_{0} < 0$ , meaning the fluid right at the wall is now creeping backward — reversed (backflow) flow. Forward flow above, backflow below → the boundary layer lifts off the surface and rolls into a wake. This feeds the form drag.

In figure s08 compare the yellow curve (separation: flat tangent at the wall) with the pink dashed curve (just downstream: the profile pokes to $u < 0$ near the wall). The blue dot marks $τ_{w} = 0$ .

The one-picture summary

Figure s09 lays the four key profiles side by side along the wall — read it left to right as a movie of the flow travelling downstream.

Everything above compresses into one march of profiles along the wall:

Favourable region ( $d p / d x < 0$ ): fat profile, steep at the wall, no inflection → attached.
Adverse region ( $d p / d x > 0$ ): the boxed law $μ \partial^{2} u / \partial y^{2} ∣_{0} = d p / d x$ forces positive wall curvature, while the curve must flatten into $U$ (negative curvature) up top → an inflection point appears → the profile thins at the base.
Separation point: the wall slope reaches zero → $τ_{w} = 0$ (the blue dot marked SEP).
Downstream: wall slope negative → backflow → the layer peels off into a wake.

Recall Feynman retelling of the whole walkthrough

A velocity profile is just a picture of "how fast is the air at each height above the wall". At the ground the air is glued still (no-slip), higher up it moves full speed. The slope of that picture at the ground is the friction — how hard the air grips the wall. We wrote Newton's law for a fluid speck inside the thin layer, then stood exactly on the wall: because the air can't slide along the wall (no-slip → speed zero) and can't pass through the solid wall (impermeability → cross-speed zero), two whole terms vanish and we're left with a tidy tug-of-war between pressure and viscosity. That gives one gem: the way the profile bends at the wall is decided purely by whether pressure is rising or falling. Falling pressure (downhill) → the profile is nice and fat, air grips hard, all attached. Rising pressure (uphill) → the wall bends one way but the top must flatten into the free stream the other way, so an S-bend with slow, easily-shoved air appears at the bottom; keep going and the ground-level slope drops to zero — that's separation — and just past it the near-wall air rolls backward and the smooth flow peels away.

Active recall

Recall Which two boundary conditions kill the convective terms at the wall, and what does each fix?

No-slip fixes the tangential speed ( $u = 0$ ); impermeability fixes the normal speed ( $v = 0$ ). Together they zero both convective terms.

Recall What sets the sign of the profile's curvature at the wall?

The sign of $d p / d x$ , via $μ \partial^{2} u / \partial y^{2} ∣_{0} = d p / d x$ . Adverse ( $> 0$ ) → positive wall curvature.

Recall Why must an adverse gradient create an inflection point?

Wall curvature is positive; but as $u \to U$ the slope decreases to zero, forcing negative curvature just below the free stream. Positive-to-negative means curvature passes through zero — an inflection.

Recall State the separation condition and what happens just downstream.

$τ_{w} = μ \partial u / \partial y ∣_{y = 0} = 0$ ; downstream the wall slope goes negative → backflow.