Research Note: Why a Bicycle Without Gyroscopes Still Refuses to Fall — Epoche B2
Research Note — Bicycle Self-Stability Topic: Testing the claim that spinning wheels keep a bicycle upright. Status: Literature review plus model check. 1. The problem Almost every textbook explanation says a bicycle stays up because its wheels act like gyroscopes. The claim rests on a real piece of mechanics. A spinning wheel has angular momentum $L = I\omega$ — the rotational counterpart of ordinary momentum, where $\omega$ is the spin rate in radians per second and $I$ is the moment of inertia, the measure of how far the wheel's mass sits from its axle. Angular momentum is a vector, pointing along the axle, and it changes only when a torque is applied, which is why a spinning top resists being tipped over. If the bicycle leans at a rate $\Omega$ — the lean angular velocity, not the lean angle — the wheel's angular momentum vector is being swung around, and that requires a torque $$\tau = \Omega \times L,$$ whose direction, by the geometry of the cross product, is perpendicular to both. For a leaning bicycle wheel this torque acts about the steering axis: the front wheel is pushed to steer into the direction of the fall. So the mechanism is genuine, and it does produce steering. The question is whether it is the mechanism that matters. An order-of-magnitude check makes the question sharper. Take typical values: a front wheel of radius $0.35\,\mathrm{m}$ with moment of inertia about its axle of roughly $0.3\,\mathrm{kg\,m^2}$, rolling at $5\,\mathrm{m\,s^{-1}}$, so $\omega = v/r \approx 14\,\mathrm{rad\,s^{-1}}$ and $L \approx 4\,\mathrm{kg\,m^2\,s^{-1}}$. A lean developing at $0.5\,\mathrm{rad\,s^{-1}}$ then gives a gyroscopic steering torque of about $2\,\mathrm{N\,m}$. Set against this the toppling torque on a bicycle and rider of $80\,\mathrm{kg}$ whose centre of mass is $1\,\mathrm{m}$ above the ground and leaning by $5^\circ$: $\tau = mgh\sin\phi \approx 80 \times 9.81 \times 1 \times 0.087 \approx 68\,\mathrm{N\,m}$. The gyroscopic effect is a few per cent of the gravitational one. It is not nothing — a small torque applied to a steering assembly of low inertia can still turn it — but the arithmetic already shows that a story resting on gyroscopic dominance is implausible. Jones (1970) made essentially this estimate and drew the same conclusion. A second popular explanation is trail , also called the caster effect. Extend the steering axis (the line through the head tube, about which the handlebars turn) down to the ground; on a normal bicycle it strikes the ground a distance $c$ in front of the tyre's contact patch, typically $5$ to $10\,\mathrm{cm}$. Because the contact point trails behind the pivot, any sideways force on the tyre generates a torque that tends to swing the wheel back into line — the mechanism that keeps a shopping-trolley castor pointing backwards, and the reason a bicycle leaned to the left will steer to the left, tightening the turn until the machine rolls back under its own weight. The problem: are either of these actually necessary? Both stories are plausible, both describe real effects, and both have been repeated for a century. An explanation should be testable, not merely plausible, and the way to test a claim of necessity is to remove the thing and see whether the effect survives. 2. The method Kooijman et al. (2011) attacked the question experimentally. They built a "two-mass-skate" (TMS) bicycle with two decisive modifications, each aimed at one of the two candidate explanations: Each wheel was paired with a counter-rotating twin, geared to spin at equal rate in the opposite sense, so the net angular momentum was $L_{\text{net}} = I\omega + I(-\omega) = 0$. Since the gyroscopic torque is proportional to $L$, cancelling $L$ cancels the torque exactly, not merely approximately. Note what this does not change: the wheels still roll, still carry load, still constrain the machine to move along their own planes. Only the gyroscopic term is deleted. The contact point was placed ahead of the point where the steering axis meets the ground: negative trail, $c < 0$. A castor with negative trail is not neutral but actively perverse — like a trolley wheel mounted backwards, it amplifies a disturbance instead of damping it. So the caster effect was not merely removed but reversed. What remained was the mass distribution. The front assembly's mass was concentrated low and forward of the steering axis, so that when the machine leans, gravity acting on that mass produces a torque about the steering axis that turns the front frame towards the lean. Formally, a leaned bicycle's potential energy depends on the steer angle, and the front assembly rolls "downhill" in that energy landscape by steering into the fall. The theoretical frame is the linearised Whipple model, first derived by Francis Whipple in 1899 and re-derived, corrected and benchmarked by Meijaard et al. (2007). It treats the bicycle as four rigid bodies — rear frame, front assembly, and two wheels — meeting at frictionless hinges, with knife-edge wheels that roll without slipping and no energy losses anywhere. Linearised means the equations are truncated to terms of first order in the small quantities, which are the lean angle $\phi$ from vertical and the steer angle $\delta$ from straight ahead; the resulting equations are exact in the limit of small deviations from upright, straight-line motion at constant speed $v$. They read $$\mathbf{M}\ddot{\mathbf{q}} + v\,\mathbf{C}_1\dot{\mathbf{q}} + \left(g\,\mathbf{K}_0 + v^2\,\mathbf{K}_2\right)\mathbf{q} = 0, \qquad \mathbf{q} = (\phi, \delta)^{\mathsf{T}},$$ which is the two-variable version of the mass-spring equation every first-year student meets, with each coefficient now a $2\times2$ matrix so that lean and steer can drive each other. Reading the terms: $\mathbf{M}$ collects the masses and moments of inertia; $v\,\mathbf{C}_1$ gathers everything proportional to speed and to a rate of change — the gyroscopic torques live here, together with inertial coupl