Push straight down, and it decides to lean

A column pinned on a rotational spring, carrying a load P straight down on top. Below a threshold, standing straight is the only stable answer. Cross it, and standing straight stops being an option — the column must lean, one way or the other, and which way comes down to the smallest of nudges.

What you're seeing

A rigid column is pinned at its base on a rotational spring of stiffness k, and carries a load P pressing straight down on its tip. If the column tilts by a small angle θ, the load's weight gets a moment arm to push it further off-centre, while the spring pulls it back toward straight. The chart plots every angle the column could sit still at, against the load. For small P there's only one such angle: θ = 0, standing straight, drawn as the solid flat line on the left.

Drag the load slider to the right and watch the flat line all the way to the amber landmark — the critical load, Pcr. Cross it, and standing straight is still technically an equilibrium (the dashed continuation), but it's no longer stable: the tiniest disturbance grows instead of fading. Two new stable answers appear either side of it, a symmetric pair of leans, +θ* and −θ* — a pitchfork, splitting from one stable answer into two. Press nudge it right as you cross the landmark and watch the sign of that nudge decide, once and for all, which tine of the fork the column takes.

Unlike yesterday's switch, this one keeps no grudge: ease the load back down below Pcr and the column returns to standing straight every time, from either lean, no loop, no memory. But watch the verdict line as you creep the load up toward the threshold without quite crossing it — recovering from a nudge gets visibly, measurably slower the closer you get, with no sudden warning beforehand. That slowing is not a story; it's the same threshold showing up in how long a disturbance takes to fade, and it diverges exactly as Pcr is approached.

The rule, exactly. The column obeys θ″ + cθ′ + kθ − Psin(θ) = 0 Linearizing near θ=0 gives an effective stiffness of (k−P): stable while P < k, unstable once P > k — so Pcr = k exactly. Past threshold, the new stable angles solve kθ* = Psin(θ*) Verified in node (improve/verify/74-buckling.js, 33 checks): simulating the exact differential equation confirms Pcr = k across four spring stiffnesses, and the simulated post-buckling angle matches an independent bisection root of that same equation to five decimal places, on both symmetric branches. Near the threshold, the time for a nudge to decay to half its size matches a closed-form linear-stability prediction and grows more than threefold as the load approaches Pcr — critical slowing down, confirmed two independent ways, not just asserted. Three negative controls confirm the test actually discriminates: a naively linearized version of the same model (sin(θ) replaced by θ) never settles at all, growing past a trillion instead of the real model's calm, bounded lean; an unsprung column (k=0) buckles under any load however small, exactly as Pcr=0 predicts; and a plausible-but-wrong guess of Pcr = k/2 is directly ruled out, since the column demonstrably stays stable at loads well above that guess.

← the cabinet · Thought Toys — a cabinet of explorable explanations. Exhibit 74.