A portal · the ground becomes a network
Steady Only in Motion
Draw the free-body diagram of an upside-down pendulum, a riderless bicycle, a cat dropped belly-up. Every one of them, held still, falls. Let them move and three of them hold a balance the still picture swears is impossible — and a fourth, spun about the wrong axis, loses one it should have kept. The motion is not decoration. It puts terms into the equations that a static diagram has no slot for, and those terms are where the balance lives.
This is a portal: it takes four pieces already on the ground and supplies the one claim none of them states alone. Each is a small, famous mechanics result — Kapitza's pendulum, the self-stable bicycle, the falling cat, the tumbling intermediate axis. Pull back far enough and they are one observation about what equilibrium even is.
The claim, in one line
A static equilibrium is a place where the forces, frozen, cancel. A dynamic equilibrium is something else entirely: a steady pattern of the motion, kept by terms that exist only because the thing is moving. Write the equation of motion and three such terms appear that a force-balance never shows you —
— and each one builds a stable state out of motion where standing still offers only a fall. The catch, and the honesty of the thing, is that the same enlargement of the rules runs the other way: motion can also take a stability away. The four demonstrations below are the proof, each recomputed in front of you.
1 · Motion digs a valley where statics has a hill
A pendulum has one resting place: straight down. Straight up is the textbook picture of instability — the faintest nudge topples it. Now vibrate the pivot up and down, fast. Past a sharp threshold a second resting place appears, pointing at the ceiling, and it pushes back when you shove it. Nothing is balancing on the buzz: the fast shaking, averaged over its own cycle, adds a real restoring force — an effective potential — that digs a valley at the top where gravity left a hill.
Left: the true equation of motion φ″ = −((g − aω²cos ωt)/L)·sinφ, integrated step by step — the √2 threshold is never used to decide anything, it only emerges. Right: the time-averaged effective potential U(θ) = −cosθ + (K²/4)·sin²θ. Watch the point at the top (θ = 180°) flip from a peak to a pocket as K crosses √2.
Drag K below 1.414 and the top falls no matter how you start it; lift it above and the bob refuses to drop. The offline verifier finds the same boundary by simulation — Kcrit ≈ 1.425, within 1% of √2, converging to it as the shake gets faster. It is the same trick that traps a single ion in a Paul trap and focuses a particle beam. Build the full pendulum →
2 · Motion supplies a feedback the rider never could
Let go of a moving bicycle and it does not fall. It wobbles, steers into the wobble, and stands back up with nobody aboard. The folklore credits the gyroscope of the spinning wheels — and the folklore is wrong: in 2011 the same researchers built a bicycle with its gyroscopic effect cancelled to zero and no trail, and it balanced itself anyway. What holds it up is the geometry of steering: when it leans, it turns, and the turn sweeps the wheels back under the falling mass. That correction exists only because the bike is rolling — it is a term proportional to speed, invisible to a standing bike.
Here is the actual benchmark machine — the reference model of the 2007 review. Drag the speed and watch the four eigenvalues of the linearized bike. Where every one has a negative real part, a riderless bike recovers; outside that band it falls.
Each dot is one eigenvalue (horizontal = growth rate, the value that matters). The shaded strip is the self-stable window the page finds by scanning, not by being told: it lands on 4.292 to 6.024 m/s, reproducing the published numbers to three decimals from the matrices alone.
Notice the right edge: ride faster than ≈6.02 m/s and the bike capsizes again. More wheel-spin, less stability — the exact opposite of the gyroscope story. Stability here is a band, not a floor. Ride the benchmark bike →
3 · Motion does what statics forbids outright
The first two manufacture a stability. The cat does something stranger: it performs a maneuver that is flatly impossible to a rigid body. Drop a cat belly-up with no spin and it lands feet-down — having turned a half-circle while its angular momentum stayed exactly zero, start to finish, with nothing to push against. A rigid body at zero angular momentum can never change its facing, ever. A body that can change shape can: circle a bent waist once and the whole body rolls over, owing the turn to geometry, not spin.
The body turn is the holonomy of a closed loop in the space of shapes — a geometric phase, the classical cousin of Berry's phase. The verifier's full two-segment model turns 179.99° for the best single circle and confirms the tell-tale of geometry: a back-and-forth wiggle (a stroke that is not a loop) turns ~0°, and reversing the loop reverses the turn.
No tail required — tailless cats land feet-down just the same; the tail is a counterweight, not the mechanism. Kane & Scher proved the half-turn in 1969 on a NASA grant. Watch the cat turn on nothing →
4 · The dark twin — where motion takes a balance away
If motion only ever gave stability, the lesson would be too tidy to trust. It doesn't. Every rigid object has three axes to spin around. Two are steady; the one in the middle is not. Spin a free body about its intermediate axis and it flips itself over, again and again, with nothing touching it — the tennis-racket theorem, the Dzhanibekov effect. It is a theorem, not a glitch, and it is the same kind of fact as the other three: a stability decided not by the static shape but by the motion. Here the motion writes a positive growth rate into the equations, and the spin runs away.
The body's facing, from a from-scratch integration of Euler's torque-free equations. Only the middle axis has a real positive growth rate σ = Ω√((I₃−I₂)(I₂−I₁)/I₁I₃); the other two give a pure imaginary value — a wobble that never grows.
The cat and the racket are two halves of one law read from opposite ends: a deformable body at zero momentum can turn itself a half-circle; a rigid body with momentum cannot help tumbling about its middle. Toss the racket yourself →
One claim, four mechanisms — and they are not the same trick
The temptation, having lined these up, is to say they are all "the same thing." They are not, and the honest portal refuses the shortcut. What they share is the claim at the top: the static picture is incomplete, and the missing piece is a term that lives in the motion. How each one does it is genuinely different — and naming the difference is the whole point.
| member | what moves | the term motion adds | the verified number |
|---|---|---|---|
| Kapitza pendulum | an external fast vibration of the pivot | a time-averaged effective potential (method of averaging) | top stable for K > √2; sim → 1.425 |
| self-righting bike | steady forward rolling | a speed-coupled feedback torque (steer-into-the-fall) | self-stable 4.292–6.024 m/s |
| falling cat | a cyclic internal change of shape | a path-dependent geometric phase (holonomy at L = 0) | ~180° per loop; ~0° per wiggle |
| intermediate axis | a steady rigid spin | a positive growth rate — the term destabilizes | σ > 0 only about the middle axis |
Averaging, feedback, holonomy, instability: four distinct chapters of mechanics, each with its own century of names — Kapitza and the ponderomotive force; the benchmark-bicycle eigenvalues; Montgomery, Shapere and Wilczek on gauge kinematics; Euler in 1758 on the asymmetric top. The portal does not collapse them. It points at the seam they all sit on: the difference between the world a free-body diagram can draw and the world the equations of motion actually run. Equilibrium, it turns out, is sometimes a verb.
The four members
Related portals on the seam of what a static picture misses: The Level and the Rate (four "facts" that are really rates), and Find What Doesn't Change (invariants that forbid the impossible). This one is their kinetic cousin: the equilibria a free-body diagram can't draw.