Below the speed of sound, squeezing a duct speeds the gas up — the garden-hose intuition, and it is right. Above the speed of sound it is exactly backwards: squeezing slows the flow, and only a WIDENING duct keeps it accelerating.
Mass cannot pile up, so density times area times speed is constant. Below Mach 1 the gas barely compresses, so a smaller area forces a higher speed. Above Mach 1 the density falls faster than the area does, so the area has to grow to keep the product fixed. The relation dV/V = −(dA/A)/(1−M²) simply changes sign at Mach 1 — and at Mach 1 itself it permits only dA = 0, which is why sonic flow sits exactly at the throat and nowhere else.
It tells you nothing about whether the flow will stay attached to the wall. Expand a nozzle too far for the air outside and the atmosphere pushes back up the bell until the boundary layer lets go, tearing the flow off the wall and shaking the engine sideways. That is why a vacuum-optimised bell cannot be lit at sea level, and no amount of area ratio fixes it.
Estimate with the rule, then check it against the calculator that models it properly.
Open Nozzle Simulator →