A converging-diverging (de Laval) nozzle's exit pressure Pe is fixed once its geometric expansion ratio ε = Ae/At is cut into the bell — it cannot change in flight. Ambient pressure Pa, however, falls roughly exponentially with altitude h:
P_a(h) ≈ P0 · exp(−h / H), P0 = 101.3 kPa, H ≈ 7.64 km
Isentropic nozzle relations (γ ≈ 1.2 for combustion gas):
A/A* = (1/M) [ (2/(γ+1))(1 + (γ−1)/2 M²) ]^((γ+1)/(2(γ−1)))
Pe/Pc = (1 + (γ−1)/2 Me²)^(−γ/(γ−1))
Thrust coefficient:
CF = √[ (2γ²/(γ−1))(2/(γ+1))^((γ+1)/(γ−1)) (1 − (Pe/Pc)^((γ−1)/γ)) ]
+ (Pe − Pa)/Pc · ε
A bell nozzle is only perfectly expanded (Pe=Pa, maximum CF) at one design altitude. Lower down it is over-expanded (Pe<Pa) — the plume necks inward and, once Pe/Pa drops below roughly 0.4, the flow risks separating from the wall before reaching the exit (Summerfield criterion). Higher up it is under-expanded (Pe>Pa) and the plume flares outward, wasting some of the pressure energy that never got converted to axial thrust.
An aerospike (plug) nozzle has no confining outer wall — the exhaust expands freely against the ambient air along the outside of a tapered spike, so the effective expansion ratio adjusts itself to track close to the ideal Pe≈Pa condition at every altitude, up to a maximum set by how far the spike is truncated. This model caps that effective ratio at 1.6× the bell's design ε (a real, compact linear-aerospike spike cannot be arbitrarily long), so the two curves converge again at very high altitude once the spike's own limit is reached — matching flight data from tested aerospike engines (e.g. the X-33 J-2T / XRS-2200 program), which showed the biggest sea-level-to-vacuum CF advantage low in the atmosphere.
- Altitude — moves the rocket up the exponential atmosphere; watch both plumes and the graph respond.
- Bell expansion ratio ε — re-designs the bell for a different target altitude; a high ε over-expands badly at sea level.
- Chamber pressure Pc — scales how much the ambient-pressure mismatch matters relative to the momentum term.