Diffusers

The question

Slowing air down should be easy: just widen the pipe. So why do engine designers deliberately choose a diffuser that throws the flow off a cliff?

What a diffuser is for

Lesson 1 left us with a clear requirement: the air arriving at ~150 m/s must be slowed to something a flame can be held in. But the diffuser is not just a brake. Its real job is to convert speed into pressure as efficiently as possible, because pressure is the currency the turbine spends to make work.

Bernoulli tells us the ideal story. For an incompressible flow in a duct whose area grows from $A_1$ to $A_2$ , continuity forces the velocity down in proportion ( $\cair{V_2} = \cair{V_1}\, A_1/A_2$ ) and the lost kinetic energy reappears as static pressure. We measure how much of the inlet dynamic pressure $q_1 = \tfrac12\rho \cair{V_1}^2$ is recovered with the , gold for recovered pressure:

$\cfuel{C_p} = \frac{p_2 - p_1}{q_1}, \qquad \cfuel{C_{p,\text{ideal}}} = 1 - \left(\frac{A_1}{A_2}\right)^2$

Whatever the real diffuser fails to recover is lost for good, as a drop in total pressure:

$\cflame{\lambda} = \frac{p_{t1} - p_{t2}}{q_1} \;\approx\; \cfuel{C_{p,\text{ideal}}} - \cfuel{C_p}$

Every bit of $\lambda$ is pressure the turbine never sees, which means fuel burned for nothing.

Why boundary layers hate slowing down

A diffuser is fighting an adverse pressure gradient: pressure rises in the direction of flow. The fast core of the flow has plenty of momentum to push "uphill" into that rising pressure. But the air right next to the wall has been slowed by friction. It has very little momentum to spare.

Diverge the walls too quickly and that tired near-wall fluid simply stops, then turns around. The boundary layer , a region of stalled, recirculating air grows along the wall, and the main flow shoots through the remaining gap as a jet. The effective area stops growing, so the velocity stops falling and the pressure isn't recovered.

Estimate first

A faired diffuser must triple its area, and the walls may diverge by only 8° in total before the flow separates. How long must it be, in inlet widths?

110100

Drag along the line to place your guess — each tick is ten times the last.

Explore the trade-off yourself. Watch the streaks: blue is fast, low-pressure air; gold is recovered pressure.

ExploreSlowing air without losing it

Area ratio A₂/A₁2.50

velocity falls to 40% of inlet

Length L / W₁8.0

Inlet distortion & swirlclean

Real compressors never deliver perfectly clean flow

Pressure recovery Cₚ0.69ideal 0.84

Total-pressure loss0.15λ, fraction of inlet dynamic pressure

Length8.0 W₁

Transitory stall: the flow is on the edge of separating. Real diffusers here flap unsteadily — bad news for the combustor downstream.

Challenges0/3

Stall a faired diffuser so badly that it recovers less than half of its ideal pressure.
Find the optimum length for an area ratio of 3 or more with a clean inlet.
Find conditions where the short dump diffuser recovers more pressure than the best faired design.

Streaks are coloured by static pressure: blue is the fast, low-pressure compressor air; gold is recovered pressure. The plot shows the faired-diffuser trade-off — too short stalls, too long drowns in friction.

Take a moment with the lower plot, because it is the whole story of faired diffusers:

Too short and the divergence angle is steep. The flow stalls, so $C_p$ collapses.
Just right means a total included angle of roughly 7–10°. The flow stays attached and recovery approaches the ideal.
Too long and friction slowly eats the recovery, while the engine grows heavier.

Then switch on inlet distortion. Air leaving a compressor is never perfectly uniform: it carries residual swirl and wakes from the blades. A distorted boundary layer is already tired, and it separates at a gentler angle. The faired diffuser's sweet spot shrinks and moves.

The dump diffuser: embracing separation

Now the surprising part. Almost every modern aero combustor uses a : a short, gentle pre-diffuser that removes some of the velocity, followed by an abrupt step, the "dump", into the space around the combustor dome.

At the step the flow does separate, deliberately. But unlike the unsteady, wandering stall on a faired wall, the separation at a sharp step is fixed in place. It forms stable corner vortices that act as soft, self-maintaining walls, and the core jet spreads into the larger space and slows. The loss of a sudden expansion from area $A_2$ to $A_3$ is the classic Borda–Carnot result, and it is modest if the pre-diffuser has already taken the speed down.

Going deeper
The Borda–Carnot sudden-expansion loss

Apply momentum conservation to a control volume spanning a sudden expansion from area $A_2$ to $A_3$ , assuming the pressure on the step face equals the jet pressure $p_2$ :

$(p_3 - p_2)\,A_3 = \rho A_2 V_2 (V_2 - V_3)$

Combined with continuity ( $V_3 = V_2 A_2/A_3$ ), the total-pressure loss is

$\Delta p_t = \tfrac12 \rho\,(V_2 - V_3)^2 \qquad\Longrightarrow\qquad C_p = 2\,\frac{A_2}{A_3}\left(1 - \frac{A_2}{A_3}\right)$

The recovery peaks at $C_p = 0.5$ when $A_2/A_3 = 0.5$ . That is only half the ideal, which is why a pre-diffuser first removes much of the velocity: the dump's loss scales with the dynamic pressure at the dump, $\tfrac12\rho V_2^2$ , not at the compressor exit.

Go back to the widget and compare: switch to the dump diffuser and crank up the distortion. The faired design's recovery falls off a cliff, while the dump barely notices.

Check your intuition

A faired diffuser performs beautifully on a clean-inlet test rig, then disappoints when bolted behind the real compressor. What is the most likely reason?

The diffuser and the combustor are one system

The dump also has a second job: it presents air to the combustor dome, which splits the flow three ways, into the swirler and the outer and inner annuli that feed the liner holes and cooling. If the diffuser delivers a lopsided or unsteady flow, the split changes: one side of the liner gets more air than the other, one set of dilution jets penetrates further. That shows up downstream as hot streaks and a poorer pattern factor (Lesson 4).

So the diffuser is never designed alone. What matters is the combined loss and uniformity of diffuser plus dome plus annuli, and the best diffuser is the one that makes the combustor's job easiest.

Recap

What to carry forward

A diffuser turns velocity into static pressure; its quality is measured by $C_p$ (recovered) and $\lambda$ (lost). Ideal $C_p = 1 - (A_1/A_2)^2$ .
Rising pressure along the flow is hard on the slow near-wall fluid; diverge too fast (beyond ~7–10° total) and the flow separates and stalls.
A faired diffuser has an optimum length: too short stalls, too long loses to friction and weight.
The dump diffuser accepts a fixed, stable separation at a step, and is short, tolerant of distortion and predictable. It is the aero-engine standard.
The diffuser, dome and annuli form one aerodynamic system whose uniformity sets up everything downstream.

Now we step through the dome, into the liner itself, to see how the swirler manufactures the recirculation zone that holds the flame, and how precisely aimed jets of air sculpt the rest.