Exploring Mechanics / 9. Unintuitive Problems
Predict first, then check — every one of these looks wrong until you work through the mechanics underneath it.
Make a prediction first: a double cone (two cones glued base-to-base) is set on two rails that spread apart as they go up a shallow incline. Released at the narrow end, it rolls toward the wide end and up the slope. Does its center of mass actually rise?
A double cone placed on diverging rails appears to roll uphill. Its center of mass actually moves downward as the rails spread apart — the shape sinks into the widening V faster than the rails climb.
These two rates are simplified stand-ins for the full 3D geometry (which depends on the incline angle, the rail half-angle, and the cone's own half-angle). Push the sink rate above the rise rate to reproduce the paradox.
No force does negative work here — the cone is simply released and rolls under gravity, converting potential energy to kinetic energy exactly as always. The "paradox" is purely about where the center of mass sits geometrically: as the rails diverge, the cone's local rolling radius at the contact points changes, and for the right combination of rail half-angle α and cone half-angle β, that radius change lowers the axle faster than the incline raises it. Energy conservation (E = mgh + ½mv² + rotational terms) is never violated — h is just decreasing the whole time.
Make a prediction first: a spool of thread rests on a table with the string unwinding from underneath the hub and running out at some angle above the floor. You pull the string toward yourself. Does the spool always roll toward you?
Pulling a string wound around a spool can make the spool roll either toward the pulling force or away from it, depending only on the angle of the string.
Moment arm about the contact point: d = Rcosθ − r · critical angle: θc = arccos(r/R)
Take torques about the ground contact point P (friction there produces no torque about itself). The string leaves the hub tangent to the circle of radius r. Working out the perpendicular distance from P to that tangent line for a string exiting underneath the hub at angle θ gives exactly Rcosθ − r. This changes sign at θc = arccos(r/R), which is why the rolling direction flips there.
Make a prediction first: a hoop, a solid cylinder, a hollow sphere, and a solid sphere — all released from rest at the same height on the same ramp. Do they reach the bottom together, the way objects in free fall do?
They do not. Even though every shape starts with the same potential energy and reaches the bottom with the same total energy, more of a hoop's energy goes into spinning rather than moving forward, so it arrives last.
a = gsinα / (1 + β) where β = I/(mR²)
Make a prediction first: a slinky hangs stretched from one end. The moment it's released, does the bottom immediately start falling, along with the rest of it?
The bottom of a stretched slinky can remain almost motionless immediately after release — it doesn't start moving until a compression wave, traveling down from the released top, physically reaches it.
This models the slinky as a chain of point masses connected by springs, starting from the correct stretched equilibrium under gravity, then releases the top and integrates the motion of every node.
A slinky is not rigid — it behaves like a very soft spring with mass distributed along its length, so disturbances propagate through it at a finite wave speed, just as sound propagates through air at a finite speed rather than everywhere at once. Because the whole slinky's center of mass has no external support once released (aside from gravity), it falls at exactly g throughout — but individual coils can still be motionless relative to the ground while the center of mass falls, as long as the coils above them are collapsing downward just as fast.
Make a prediction first: a load hangs from a movable pulley, supported by a rope that runs up to a fixed pulley and back down to your hand. If you pull one meter of rope through your hand, does the load rise by one meter?
A load lifted by a movable pulley rises only part of the distance that the free end of the rope is pulled — because two rope segments support it, each one only needs to shorten by half the total distance pulled.
Rise = pull distance / 2 · Force needed = Weight / 2
This is the same mechanical-advantage tradeoff from Section 5's Pulleys tab — reduced force always costs you distance, in exact proportion.
Make a prediction first: a long bead chain is poured out of a raised container into a lower one. Does the falling chain just drop straight down toward the lower container, or could it visibly arc upward above the container's rim first?
A chain falling from a container can rise above the rim before descending — the "chain fountain" or Mould effect. The full explanation is still an active topic of physics research.
Biggins and Warner modeled the pile of chain as briefly rigid near where a link is picked up, so lifting a link there requires an additional force of roughly Fextra ≈ μv², where μ is mass per unit length and v is the chain's speed — on top of the weight it's already supporting. That extra push from the pile is what launches the chain fountain above the rim; a perfectly flexible idealized chain would show no such effect.
Make a prediction first: a bead slides frictionlessly from point A down to a lower point B, under gravity only. Is the straight line between them the fastest path?
The straight-line path is not the fastest path between two points under gravity — the ideal curve (found by Johann Bernoulli in 1696) is a cycloid, which dips down steeply at first to build speed early, even though it's a longer path overall.
All four beads start at rest at the same point and finish at the same point. Watch which one wins.
Minimizing total travel time T = ∫ ds/v with v = √(2gy) (from energy conservation) is a calculus-of-variations problem. Applying the Euler–Lagrange equation to this functional produces a differential equation whose solution is exactly the parametric cycloid x = R(θ − sinθ), y = R(1 − cosθ) — the curve traced by a point on the rim of a rolling circle of radius R.
Drag the orange control point to reshape your own ramp and race it against the true cycloid optimum — no matter how you bend it, the cycloid (dashed green) always wins or ties.
Make a prediction first: four beads are released from rest at four different points along the same curved dip, some closer to the bottom than others. Does the bead that starts closer to the bottom always arrive first?
A bead sliding on an inverted cycloid reaches the bottom in exactly the same time, no matter where along the curve it starts — the same curve that solves the brachistochrone problem is also "tautochrone" (same-time).
Time to the bottom = T/4 = π√(R/g), independent of the starting point
Christiaan Huygens proved this isochronism and built pendulum clocks with cycloid-shaped guides ("cheeks") near the pivot, forcing the bob's path to be a cycloid rather than a circular arc — making the clock's period genuinely independent of swing amplitude, rather than only approximately so for small angles as with an ordinary pendulum (compare Section 8's Pendulums tab).
Three beads, released from three very different heights on the same cycloid, running forever — watch them keep meeting at the bottom in lockstep, over and over.
Make a prediction first: a chain or cable is hung freely between two supports, sagging under its own weight. Is the resulting curve a parabola, the same shape a thrown ball traces?
A freely hanging uniform chain forms a catenary, not a parabola — a subtly different curve described by a hyperbolic cosine, even though the two look nearly identical for small sag.
Catenary: y = a·cosh(x/a) − a vs. parabola: y = x²/(2a)
Balancing horizontal tension T0 against the weight of chain hanging below each point gives the differential equation d²y/dx² = (1/a)√(1+(dy/dx)²), with a = T0/(weight per length). Solving it produces y = acosh(x/a) − a, exactly. Expanding cosh in a Taylor series, cosh(u) ≈ 1 + u²/2 for small u, shows why the catenary looks almost exactly parabolic near its lowest point but visibly diverges near the supports.
Make a prediction first: a rigid object (a tennis racket, a book held closed with a rubber band, a smartphone) has three different principal axes of rotation. Thrown spinning cleanly about any of the three, does it keep spinning cleanly about that same axis?
Rotation about an object's intermediate principal axis is unstable, causing unexpected flipping mid-flight — the intermediate-axis theorem, also called the Dzhanibekov effect after the cosmonaut who noticed it with a spinning bolt in orbit.
I1dω1/dt = (I2−I3)ω2ω3 (and cyclic permutations) — Euler's free rigid-body equations
Choose which axis the object starts spinning about (with a tiny nudge on the other two, as any real throw has). Watch the three spin-component graphs — stable axes barely wobble; the intermediate axis produces huge swings and repeated flips.
Linearizing Euler's equations around ω≈(ω0,0,0) (spin purely about axis 1) gives small perturbations that oscillate, since (I1−I2) and (I3−I1) have opposite signs relative to needed for growth — same for axis 3. But linearizing around spin purely about axis 2, the corresponding coefficients both come out with signs that reinforce growth instead of oscillation, producing exponential runaway until the nonlinear terms take over and flip the object end over end.
Make a prediction first: a fast-spinning wheel is suspended by a string tied to just one end of its axle, with the other end sticking out horizontally, unsupported. Gravity pulls down on the unsupported end. Does that end simply fall?
A torque applied to a spinning wheel produces motion in a direction that doesn't match naive intuition — instead of falling, the axle sweeps around slowly in a horizontal circle (precesses).
Ω = τ/L
Torque changes angular momentum: τ = dL/dt. Because the wheel's angular momentum vector L already points along the spin axis and is large, a torque perpendicular to L (from gravity) doesn't add to L's magnitude — it rotates its direction instead, exactly the way a force always perpendicular to a velocity produces circular motion rather than a speed change. The axle sweeps sideways because that's the direction L is being pushed, not because gravity's downward pull vanishes.
This is the same angular-momentum-conservation idea from Section 7's Angular Momentum and Gyroscopes tab, isolated here as its own "surprise."
Make a prediction first: a bead sits at the bottom of a circular hoop that's free to rotate about a vertical diameter. As you spin the hoop faster and faster, does the bead just stay put at the bottom the whole time?
As hoop rotation increases past a critical speed, the stable equilibrium position of the bead splits into two new positions away from the bottom — a textbook example of a bifurcation.
Critical rotation speed: Ωc = √(g/R) · new equilibria (for Ω > Ωc): cosθ = g/(Ω²R)
In the rotating frame, the bead behaves as if moving in an effective potential Veff(θ) that combines real gravity with the centrifugal term: Veff ∝ −cosθ − (Ω²R/2g)sin²θ. For small Ω, this potential has a single minimum at θ=0. Past Ωc, that minimum turns into a local maximum (unstable), and two new minima appear symmetrically — a pitchfork bifurcation, the same mathematical structure that shows up across many branches of physics wherever a symmetric equilibrium loses stability as a parameter is increased.