The speed that does not come back, and the √2 that separates it

Gravity has infinite range. It never switches off, and there is no distance at which a body has finished escaping. So the phrase “escape velocity” ought to be meaningless — whatever speed something leaves at, gravity is still pulling on it.

It is not meaningless, because the total pull over an infinite distance is finite. The potential well has a definite depth, and a body arriving with more kinetic energy than that depth will still be moving when the pull has run out.

The energy statement #

Set the total energy to zero — the boundary between bound and unbound:

$$\frac{1}{2}v^2 - \frac{GM}{r} = 0 \quad\Longrightarrow\quad v_{\text{esc}} = \sqrt{\frac{2GM}{r}}.$$

The circular speed at the same radius comes from balancing gravity against the centripetal requirement, $v_{\text{circ}} = \sqrt{GM/r}$. Divide one by the other and everything cancels:

$$\frac{v_{\text{esc}}}{v_{\text{circ}}} = \sqrt{2}.$$

No mass, no radius, no constants. The ratio is $\sqrt{2}$ at every distance from every body in the universe, and the gap visible in the figure is the same gap everywhere.

The consequence for spaceflight is encouraging. Reaching low Earth orbit takes 7.8 km/s; leaving Earth entirely takes 11.2 km/s. Getting into orbit is 71% of the speed needed to leave the planet, which is the origin of the observation that low orbit is halfway to anywhere. In energy terms it is exactly half, since energy goes as the square.

The direction that does not matter #

A curious feature of the escape condition: it says nothing about direction.

Energy is a scalar. A body launched at 11.2 km/s straight up escapes; so does one launched at 11.2 km/s at 45°, or horizontally, or at any angle that does not run into the ground. The trajectories are wildly different — one radial, one a wide parabola — and all of them are unbound.

That is a real consequence of energy conservation and it is worth checking against intuition, which insists that going up must be more effective. It is not. What direction changes is the shape of the escape trajectory and the time taken, not whether escape happens.

There is a strong practical caveat. Launching horizontally at ground level means traversing the atmosphere at 11.2 km/s, which is not survivable. Real launches go up first to clear the air and then turn, and the turn is a compromise between drag and gravity losses rather than anything to do with escape. Escape speed produces a parabola, which as a member of the conic family is the case of exactly zero energy. A body on it arrives at infinity with exactly zero speed — it escapes, and only just. Any excess puts it on a hyperbola, arriving with a residual speed called $v_\infty$, and that residual is the quantity interplanetary mission design actually cares about.

The numbers, and what they decide #

The last two rows matter more than they look. Leaving the solar system from Earth’s orbit requires 42.1 km/s relative to the Sun, and Earth already provides 29.8 of it. A spacecraft that escapes Earth in the right direction inherits that, and needs only the difference — which is why every outer-planet mission launches along Earth’s orbital motion, and why a gravity assist is worth so much.

Escape speed also decides which atmospheres survive. Gas molecules have a distribution of speeds, and the fastest tail of that distribution can exceed escape speed and leak away. Since lighter molecules move faster at a given temperature, they leak first. The Moon has 2.4 km/s and no atmosphere. Earth has 11.2 and has kept nitrogen and oxygen while losing essentially all its hydrogen and helium. Titan is smaller than Mars with an escape speed of 2.6 km/s and has a thick atmosphere, because it is cold enough that the tail never reaches. The composition of a planet’s air is a competition between escape speed and temperature.

The deeper the well, the more a burn is worth #

Escape speed rises steeply as a body is approached, and that steepness makes the timing of a burn matter enormously.

A burn adds a fixed amount of velocity. The energy it adds is not fixed — the change in kinetic energy is $\Delta(v^2/2)$, which for a given $\Delta v$ grows with the speed the spacecraft already has. Burning deep in a gravity well, where the spacecraft is moving fastest, converts propellant into orbital energy far more efficiently than burning far out.

This is the Oberth effect, and it is the reason a departure burn is made at the lowest point of the orbit rather than at leisure. Parker Solar Probe uses it in reverse: each Venus flyby lowers its perihelion, so each subsequent burn happens deeper in the Sun’s well and does more.

The effect has an unintuitive corollary. To reach the Sun is harder than to leave the solar system. Earth’s 29.8 km/s of orbital motion has to be cancelled to fall inward, against 12.3 km/s to escape outward from Earth’s orbit. Getting to the Sun is more than twice as expensive as getting to interstellar space, which is why the Parker probe needed seven Venus flybys and why nothing has ever simply flown at the Sun.

Where it stops being a speed #

Push the formula to its limit and it produces the strangest object in astrophysics.

Escape speed grows as the square root of $M/r$, so a body compact enough will have an escape speed exceeding the speed of light. Setting $v_{\text{esc}} = c$ gives

$$r_s = \frac{2GM}{c^2},$$

which for the Sun is 3 km and for the Earth about 9 mm.

Michell in 1783 and Laplace in 1796 both did this calculation and both concluded that sufficiently massive stars would be invisible. Their reasoning was Newtonian and their formula is right, which is a coincidence: the correct general-relativistic derivation gives the identical expression for entirely different reasons. The Newtonian picture has light slowing down and falling back like a thrown ball, which is wrong; the relativistic one has the geometry of spacetime closing off every outward path, which is not a statement about speed at all.

At that point “escape velocity” stops being a useful concept. Nothing escapes, not because it would need to move faster than light, but because there is no outward direction to move in.

What was actually measured #

Nothing in the table above was measured. Escape speed is not an observable quantity; it is computed from two numbers, and it is worth saying where each of them comes from and how well it is known.

The first is $GM$, the gravitational parameter of the body. For the Earth it is known to about nine significant figures — $3.986004418\times10^{14}\,\text{m}^3\text{s}^{-2}$ — because satellites have been tracked by laser ranging for decades and their orbits are exquisitely sensitive to it. For Mars it comes from the orbits of spacecraft that have been there; for Jupiter, from the Galilean moons, whose periods and separations have been measured since Galileo and refined by every flyby. For an asteroid visited by nothing, $GM$ is inferred from a size estimate and an assumed density, and may be wrong by a factor of two.

The second is $r$, and the ambiguity in it is larger than it looks for anything that is not a rigid sphere. “Escape speed from Jupiter” requires deciding what Jupiter’s surface is, and the convention — the one-bar pressure level — is a choice rather than a boundary. The 59.5 km/s in the table is the escape speed from a radius nobody could stand on.

The consequence of escape speed, on the other hand, is measured directly and continuously. Atmospheric escape is an observed flux, not an inference. Earth loses hydrogen at roughly 3 kilograms per second, measured from the extended hydrogen corona seen in Lyman-alpha and from in-situ ion measurements over the poles. Mars loses about 2 to 3 kilograms per second, measured by the MAVEN spacecraft, which flew through the escaping population and counted it — and which found that the rate spikes by an order of magnitude during solar storms, so the integrated loss over four billion years depends on the history of the Sun’s activity rather than on any property of Mars alone.

That last point is the honest version of the tidy story in the previous section. Thermal escape, the mechanism the speed distribution describes, is not the dominant loss channel for Mars now and probably never was. What removes a planet’s air is mostly the solar wind stripping ions from an atmosphere with no magnetic field to shield it — a process in which escape speed appears as a threshold but not as the driver.

The same two curves are worth reading at three very different ranges, because the ratio between them is fixed and everything else about the picture is not.

The same ratio, at the scale of a galaxy #

The $\sqrt{2}$ looks like a fact about spacecraft. It is a fact about any system held together by an inverse-square force, and following it upward gives one of the strongest arguments in astronomy for something that has never been seen.

The generalisation is the virial theorem: for a self-gravitating system in equilibrium, the time-averaged kinetic energy is exactly half the magnitude of the potential energy. The escape-speed relation is that theorem for a single orbiting particle. Applied to a swarm of them — a star cluster, a galaxy, a cluster of galaxies — it says the same thing about the swarm’s velocity dispersion, and it turns a set of measured speeds into a mass.

The measurement runs the argument backwards. Measure how fast the members are moving; ask what mass would be required to keep them bound; compare with the mass that is visible. Fritz Zwicky did this for the Coma cluster in 1933 and found the galaxies moving far too fast for the light he could see to hold them — by a factor of several hundred on his numbers, and by a factor of about five or six on modern ones. His conclusion was that the cluster contained a great deal of matter that did not shine, and he called it dunkle Materie. It was ignored for forty years.

The same calculation at the scale of a single galaxy gives the same answer. A star at the edge of the Milky Way, 20 kiloparsecs out, is moving at roughly 200 km/s. The escape speed at that radius from the mass that can be seen is about 130 km/s. Every such star should have left, and the galaxy should have dispersed long ago.

The surprising thing is which part of the argument is doing the work. Nothing about it requires knowing what the missing mass is, or where exactly it sits, or how it behaves. It requires only that gravity go as the inverse square and that the system be old enough to have settled — and the conclusion follows from a ratio derived, in the first section of this essay, by dividing one square root by another.

Just below the line #

The interesting region is not above the escape curve but just under it, where an orbit is bound and only barely.

Long-period comets live here, and the margin is tiny. A comet with an aphelion at 50,000 AU is bound by an energy difference of about one part in $10^5$ of its perihelion kinetic energy — small enough that a passing star, or a nudge from Jupiter, can push it either way. The boundary between returning and leaving is crossed routinely by objects that were never far from it.

That convergence is a practical difficulty as well as a curiosity. Determining whether a newly discovered object is bound requires measuring an energy that is nearly zero, from an arc of orbit that is nearly a parabola whatever the answer, which is why the interstellar visitors needed weeks of tracking before anybody would say so.

The formula also treats the body as still, and a rotating one hands part of the escape speed over free.

A launch site at latitude $\varphi$ is already moving eastward at $464\cos\varphi$ metres per second from the Earth’s rotation. Launching east collects it. At the equator that is 0.46 km/s against an 11.2 km/s escape requirement — 4%, which sounds small and is worth roughly a tenth of a stage once the exponential of the rocket equation is applied.

That is why launch sites cluster near the equator and why almost everything launches east. Kourou at 5.2° north gives 0.46 km/s; Cape Canaveral at 28.5° gives 0.41; Baikonur at 46° gives 0.32. The differences are a few percent of the budget and a substantially larger percentage of the payload.

The exception proves the rule. A satellite bound for a polar or Sun-synchronous orbit must launch north or south, collects nothing, and pays a rotation penalty instead — which is why polar launch sites exist at high latitudes where the loss is smallest.

The speed a molecule needs #

The same threshold decides something that has nothing to do with spacecraft: whether a body keeps an atmosphere.

A gas at temperature $T$ has a distribution of molecular speeds whose typical value is $\sqrt{3kT/m}$, so a light molecule moves faster than a heavy one at the same temperature. Molecules in the thin upper atmosphere, where collisions are rare enough that one heading outward will not be stopped, escape if they exceed the local escape speed — and the distribution has a tail, so some always do.

The useful rule is that the loss is slow enough to be irrelevant over the age of the solar system if the typical thermal speed is below about a sixth of the escape speed, and catastrophic if it is above about a third. Between those, a planet loses an atmosphere over hundreds of millions of years.

That single ratio sorts the solar system. The Earth, at 11.2 km/s escape and an exospheric temperature of around 1,000 K, retains nitrogen and oxygen comfortably and loses hydrogen and helium steadily — which is why the atmosphere has essentially none of either despite both being abundant everywhere else. The Moon, at 2.4 km/s, retains nothing. Titan, at 2.6 km/s, retains a thick nitrogen atmosphere because it is at 94 K and the thermal speeds are correspondingly small.

Escape speed is a property of the body and thermal speed is a property of the molecule, so the same threshold applied to different gases gives a different answer for each, and a planet’s atmospheric composition is partly a record of which side of this line each species fell on.

Where the model stops #

Point masses. The formula uses the distance from the centre, which is only right outside a spherical body. Inside one, the escape speed is smaller and the calculation needs the mass enclosed.

One body. Escaping the Earth is not escaping the Sun. Every escape is from a specific well, and the solar system’s structure is a set of nested wells with boundaries where one dominates. A spacecraft on a hyperbola with respect to Earth is usually still on a perfectly ordinary ellipse with respect to the Sun.

No atmosphere. Drag makes the ballistic calculation irrelevant for any launch from a body with air.

Impulsive departure. A continuously thrusting spacecraft never needs escape speed at any moment — it can spiral out arbitrarily slowly. Escape speed is the requirement for an unpowered coast, not for leaving.

Newtonian gravity. As above, at the extreme.

The figure has a limitation shared by every plot of this kind: it shows speeds in units of the surface circular speed, which makes all bodies look the same. That is the point — the $\sqrt{2}$ is universal — and it also hides that the actual numbers span three orders of magnitude between a small asteroid and the Sun. Universality on one axis is bought by suppressing everything on the other.

Two more readings of the same pair of curves close the argument, one at each extreme.

One more reading covers the range between a useful orbit and the edge of the planet’s domain, which is where the two curves come closest to each other without ever touching.

The ladder from here #

Later rungs: gravitational potential and potential energy, and the sign conventions that make them confusing. Escape from a rotating body, and why equatorial launch sites help. Atmospheric escape mechanisms — Jeans, hydrodynamic, and non-thermal — and which planets lost what. The Oberth effect worked through. $v_\infty$ and the characteristic energy $C_3$. Spheres of influence and patched conics. The rocket equation, which converts these speeds into propellant mass and delivers the bad news. And the Schwarzschild radius done properly, where the Newtonian coincidence stops.

Michell’s 1783 paper describes objects whose light cannot escape and suggests they might be detected by their effect on companion stars. That is exactly how the first stellar-mass black hole was found, 188 years later.