The Verification Venue · pointed at a thing everyone gets backwards
The Weight That Sways So the Tower Won't
A 660-tonne steel ball hangs inside the top of Taipei 101 to keep the skyscraper from making people seasick in a typhoon. Everyone says it blocks, braces, or soaks up the sway. It does the opposite: it deliberately swings harder than the building and a quarter-cycle out of phase, so its inertia hauls back exactly when the tower lurches forward.
Here is the part no paragraph can show you. There is exactly one way to tune that pendulum so it works best — and it is a closed-form equation discovered by J. P. Den Hartog in the 1940s. Tune it wrong and you don't just lose the benefit; you make the sway worse. Below is a real coupled oscillator — a building and a hanging damper. Push the building toward resonance, then tune the damper and watch the single deadly peak split into two. Find the tuning that makes them equal. That's the optimum.
Worst-case sway
—×
tallest peak, vs. a static push
Ball vs. building (at resonance)
—×
phase —°
Two peaks
— / —
at the invariant points
Taipei 101's ball is roughly 0.2–0.3% of the structure; here we exaggerate so the effect is visible. Bigger ball, gentler the best-case sway.
How fast the ball wants to swing, relative to the tower. The whole game is here. Den Hartog says the best value is 1/(1+μ).
Too little and a sharp spike survives; too much and the ball can't swing freely, so it stops helping. The best value is √(3μ / 8(1+μ)³).
The curve is the frequency response of the coupled system: along the bottom is the frequency of the wind gusts (1.0 = the tower's own resonance); the height is how far the tower sways, as a multiple of the gentle push it would give in a steady breeze. With no damper, that height runs off to infinity right at resonance — the spike that makes a tower unliveable. Drop in a damper and the spike splits in two, because the building and the ball now resonate as one two-part system with two slightly different natural frequencies. Tune badly and one of the twin peaks towers over the other. There is one tuning where they are exactly equal — and that equal height is the lowest the worst peak can possibly be.
The check — every number recomputed in front of you
For the mass ratio μ = 0.050 you've set, Den Hartog's closed-form optimum, computed live:
ζ_opt = √( 3μ / 8(1+μ)³ ) = 0.127
peak = √(1 + 2/μ) = 6.40 × the static sway, at both peaks
At the optimum the response curve passes through two invariant points — fixed heights that don't move no matter how you set the damping — and the optimal tuning makes both peaks land on them at equal height. Recomputed for your μ, the invariant points sit at:
And right now, with your slider settings (f = 0.952, ζ = 0.127), the two peaks the curve actually reaches are:
The whole curve is the exact dimensionless main-mass response |X·K/F₀| of the 2-DOF absorber system, evaluated point by point — the same function checked offline. Run it yourself: node research/tuned-mass-damper/verify-tuned-mass-damper.mjs.
The real ball
Taipei 101's damper is a sphere of 41 welded steel plates, each 125 mm thick, 5.5 m across, weighing about 660 tonnes. It hangs on 92 cables roughly 42 m long, slung between floors 87 and 92, with hydraulic dampers at its base. During Typhoon Soudelor on 8 August 2015 it recorded a 1 metre swing — its largest ever. These are sourced figures, each shown against its reference below; the geometry checks out (41 × 125 mm ≈ 5.1 m, close to the 5.5 m sphere; a solid steel sphere that size would weigh about 680 t, consistent with 660 t of welded plate).
| Figure | Value | Source |
|---|---|---|
| Damper mass | ≈ 660 t | Taipei 101 (Wikipedia) |
| Steel plates | 41 × 125 mm | Tuned mass damper (Wikipedia) |
| Sphere diameter | 5.5 m | Taipei 101 (Wikipedia) |
| Suspension cables | 92, ≈ 42 m | Tuned mass damper (Wikipedia) |
| Location | floors 87–92 | Taipei 101 (Wikipedia) |
| Soudelor swing | 1 m, 8 Aug 2015 | Typhoon Soudelor (Wikipedia) |
What's exactly true here, and what's idealised
Exactly true. The frequency-response curve is the exact closed-form magnitude of the main mass in a two-degree-of-freedom (building + absorber) oscillator under harmonic forcing. Den Hartog's optimum f = 1/(1+μ), ζ = √(3μ/8(1+μ)³) really does equalise the two peaks: it is the classic invariant-points (fixed-point) result, and the equal height is exactly √(1+2/μ). The verifier confirms both invariant points are independent of the damping, that the optimal tuning makes them equal, and that detuning raises the worst peak — all recomputed, not asserted. The phase fact is real too: at the tower's own resonance the optimally-tuned ball swings several times further than the building and roughly 90° behind it.
Idealised — and this matters. Den Hartog's closed form assumes the primary structure has no damping of its own and that the wind forces it at a single, steady frequency. Real towers have a little inherent structural damping, and real wind and earthquakes are broadband and gusty — so the exact optimum shifts slightly, and modern designs solve the tuning numerically rather than from this formula. The formula is the right idea and the right ballpark; it is not the literal final tuning of any specific building. We have not tried to model Taipei 101's actual modal masses or frequency — the slider's μ is deliberately exaggerated so the split is visible on screen.
What a damper does and doesn't do. It reduces sway — Taipei 101's design cites up to roughly a 40% cut in peak acceleration — and its job is occupant comfort: stopping the slow, nauseating sway that makes the top floors feel like a ship. It does not stop the building moving, and it is not what keeps the tower from collapsing (that's the structure itself). The ~40% figure and the 1 m Soudelor swing are quoted from the sources, not recomputed here; the live numbers are the Den Hartog response and its optimum.