De Laval nozzle geometry — conical, bell (Rao contour), 80% bell: Question bank

This is the misconception-hunting companion to the parent topic. Every item below is a one-line reveal: read the prompt, commit to an answer out loud, then check. The answers give reasoning, never a bare "yes/no" — because the trap is always in the why.

The picture and symbols we lean on

Before any trap, we build every symbol on a drawing so nothing is assumed. Figure 1 is a straight-walled conical nozzle cut down the middle; Figure 2 is a bell nozzle drawn the same way so you can compare shapes side by side.

Figure — De Laval nozzle geometry — conical, bell (Rao contour), 80% bell

Reading Figure 1 left to right:

The axis (dashed gray) is the centreline — the direction thrust pushes.
The narrowest point is the throat; its radius is $R_{t}$ (green). This is where the gas hits the speed of sound.
The wide opening is the exit; its radius is $R_{e}$ (orange).
The straight distance from throat to exit along the axis is the length $L$ (blue bracket).
The wall rises at a fixed half-angle $α$ (red) — the angle between the sloping wall and the axis. "Half" because the full opening angle of the cone is $2 α$ .

Two facts fall straight out of this triangle, and we derive them so no formula is quoted from thin air:

Now the exit-flow angle $θ_{e}$ : it is the angle between the gas velocity and the axis at the moment the gas leaves. In a cone the gas follows the wall, so $θ_{e} = α$ . In a bell (Figure 2) the wall bends back near the exit, so the gas leaves nearly straight and $θ_{e}$ is small — that is the whole point of the curve.

Where the divergence factor $λ$ comes from

We claimed $λ = \frac{1 + c o s α}{2}$ ; here is why, tied to Figure 3.

With the picture and both formulas earned, the traps below are safe to spring.

True or false — justify

A conical nozzle with half-angle

α = 0

would have

λ = 1

True — with

λ = \frac{1 + c o s 0}{2} = 1

, but

α = 0

means a straight tube that never expands, so it is useless for producing supersonic exhaust; the ideal factor is real, the geometry is degenerate.

A bell nozzle can reach

λ = 1

exactly

False in practice — a real bell gets

λ \approx 0.98

–

0.995

because turning the flow fully axial without any shock over a finite length is impossible; only an infinitely long ideal contour reaches exactly

1

An 80% bell is 80% as efficient as a full bell

False — the "80%" refers to length (

L = 0.8 L_{cone}

), not efficiency; its

λ \approx 0.985

is nearly as high as a full bell's

\approx 0.99

The "full Rao contour" (100% bell) has the same length as its equivalent cone

True — a "100% bell" means its length

L

equals the length of a

1 5^{\circ}

cone of the same

ϵ

; an 80% bell is

0.8

of that same reference length, so "%" always measures length against the reference cone, not against a longer bell.

Increasing the cone half-angle

α

always shortens the nozzle

True for length but with a cost — larger

α

means

tan α

grows so

L = R_{t} (ϵ - 1) / tan α

shrinks, yet

cos α

falls so divergence loss rises; length and efficiency pull in opposite directions.

Two nozzles with the same expansion ratio must have the same exit velocity

False — same

ϵ

fixes the ideal Mach number, but wall shape sets how much is lost to divergence and shocks, so a bell delivers more usable axial velocity than a cone of equal

ϵ

A bell nozzle is unconditionally lighter than a conical one of the same

ϵ

and efficiency

False as a blanket rule — the bell's shorter wall saves mass, but a curved contour often needs thicker walls, stiffening rings, or heavier fabrication than a simple straight cone, so the net mass depends on structural and manufacturing choices, not length alone.

The divergence factor

λ

depends on the gas properties

False for the geometric factor —

λ = \frac{1 + c o s α}{2}

depends only on exit-flow angle; gas properties affect

v_{e}

and

\overset{m}{˙}

, which multiply

λ

but do not change it.

Spot the error

"A 60% bell is best because it's the most compact."

The error ignores the

4%

thrust loss (

λ \approx 0.96

); compactness alone is not the objective — the 80% bell is standard because it keeps

\approx 98.5%

efficiency while still being

20%

shorter than a cone.

"The exit angle of an 80% bell is

0^{\circ}

, so it has zero divergence loss."

A real 80% bell exits at

θ_{e} \approx 5

–

1 0^{\circ}

, not

0^{\circ}

; the flow is nearly axial, giving a small residual loss — only the theoretical full Rao contour aims for exactly axial exit.

"Longer nozzle → lower exit pressure → always more thrust."

Only until exit pressure equals ambient; past that the nozzle over-expands, internal shocks form, and thrust drops.

λ = \frac{1 + c o s α}{2}

can be applied directly to a bell nozzle using its exit angle."

The formula was derived (see Figure 3) for uniform conical flow across the whole exit; a bell's flow angle varies across the exit plane, so this is an approximation, not an exact result — the real

λ

comes from integrating the actual contour.

"Rao's contour minimizes nozzle length."

Rao's calculus-of-variations problem maximizes thrust for a fixed length, not the reverse; length is the constraint, thrust is what's optimized.

"Because the bell is contoured, manufacturing tolerance doesn't matter."

The opposite — a rough or mis-shaped bell wall adds friction and can create weak shocks, so a poorly made bell can underperform a smooth cone; tolerance matters more here.

"A cone with

α = 1 5^{\circ}

loses

15%

of its thrust to divergence."

No — the loss is

1 - λ = 1 - \frac{1 + c o s 1 5 ^{\circ}}{2} \approx 1.7%

, not

15%

; the angle is not the percentage, the cosine averaging (Figure 3) makes the loss much smaller.

Why questions

Why does a bell nozzle turn the flow axial near the exit instead of letting it fan out?

Sideways momentum produces no forward thrust and can even cancel across the exit; bending the wall back toward the axis (Figure 2) converts that would-be-wasted momentum into axial push, raising

λ

toward

1

Why does Rao's contour expand rapidly right after the throat?

Fast area growth near the throat drops the pressure quickly while the gas is still dense, extracting most of the available energy early — then the wall can turn gently at the end to avoid shocks.

Why must the final turning of a bell be gradual?

A sharp turn would compress the supersonic flow abruptly, creating a shock wave that dumps kinetic energy into heat and disorder, costing exactly the thrust the contour is trying to save.

Why is the conical factor an average of

cos θ

rather than just

cos α

The exhaust leaves across a whole conical surface at angles from

0

up to

α

, so you average the axial fraction over that surface (Figure 3); weighting by the ring area

2 π R sin θ

yields the mean factor

\frac{1 + c o s α}{2}

Why do vacuum-stage engines use longer nozzles than sea-level engines?

In vacuum there is no ambient pressure to over-expand against, so a larger

ϵ

(hence longer nozzle) keeps expanding the gas for more velocity.

Why doesn't a higher expansion ratio alone guarantee better real thrust?

ϵ

sets the ideal expansion, but if it over-expands for the current altitude, shocks and flow separation appear and net thrust falls; the geometry must match both altitude and wall shape, not just chase large

ϵ

Why is the 80% figure so common across engines rather than 75% or 85%?

It sits near the knee of the diminishing-returns curve — roughly the first

80%

of length captures

\approx 98%

of the efficiency, so beyond it you add weight for tiny gains; it is a practical sweet spot, not a physical constant.

Edge cases

What happens to

λ

as the cone half-angle

α \to 9 0^{\circ}

cos 9 0^{\circ} = 0

λ \to \frac{1}{2}

— a nozzle spraying gas sideways keeps only half its momentum axial; this shows the formula degrades smoothly and never predicts nonsense.

What does the length formula give when

ϵ = 1

L = R_{t} (1 - 1) / tan α = 0

— exit area equals throat area, there is no diverging section at all, which is exactly the boundary between converging-diverging and a plain throat.

What if two designers pick different

α

but demand the same exit radius

R_{e}

They get the same

R_{e} = R_{t} ϵ

but different lengths (steeper

α

→ shorter) and different

λ

(steeper → lossier); same exit size does not mean same performance.

Does all of this geometry still apply if the flow in the throat is not choked?

No — every formula here assumes the throat has reached sonic speed (choked) so the diverging section runs supersonic; at very low chamber pressure or a heavily throttled small engine the flow can stay subsonic, in which case the diverging cone simply slows the gas like an ordinary diffuser and produces no supersonic expansion at all.

When a nozzle is over-expanded enough for the flow to separate from the wall, does the geometric

λ

still describe the loss?

No —

λ

assumes the gas fills the nozzle and follows the wall to the exit; once the flow separates, the effective exit angle and area change unpredictably, so the clean

\frac{1 + c o s α}{2}

picture no longer holds.

Is a "0% bell" meaningful?

No — zero length means no diverging section, so the gas never reaches supersonic exit conditions; the length fraction must be positive for the nozzle to function at all.

Recall Harder application check

A designer halves the cone half-angle $α$ while keeping $ϵ$ fixed. What happens to length and to divergence loss, and why can't they win on both? ::: Length $L = R_{t} (ϵ - 1) / tan α$ roughly doubles (smaller $tan α$ ), while divergence loss $1 - \frac{1 + c o s α}{2}$ falls toward zero — you buy efficiency with length and weight, so the two goals trade off and no single $α$ optimises both. Two engines share $ϵ$ and $λ$ ; one is an 80% bell, one a cone. Is the bell guaranteed lighter? ::: Not guaranteed — its shorter wall helps, but thicker curved walls, stiffeners, or costlier fabrication can erase the saving; length is only one term in the mass budget.

The picture and symbols we lean on

Where the divergence factor λ comes from

True or false — justify

Spot the error

Why questions

Edge cases

Where the divergence factor $λ$ comes from