The Repository Root Cause

The Millennium Bridge Wobble Was Never About People Walking in Step

August 18, 2026 · Runtime 8:51 · Watch on YouTube

London's Millennium Bridge opened on 10 June 2000, swayed sideways by around 70 mm under the crowd, and closed two days later for twenty months. The famous explanation is that pedestrians fell into step with each other and with the deck. The engineers' own measurements say something narrower and stranger: the force the crowd fed into the bridge followed the deck's velocity, not its position — which is damping with the sign reversed.

Chapters

  1. 0:00 Open on Saturday, shut on Monday
  2. 0:56 A bridge built to look like it is not there
  3. 1:27 Opening day, in measured quantities
  4. 1:59 The explanation everybody repeats
  5. 2:34 Arup puts people on the bridge
  6. 3:09 The force followed the speed
  7. 3:43 The critical number of pedestrians
  8. 4:15 2021: synchronisation is not required
  9. 4:54 Why the distinction is not semantics
  10. 5:29 37 viscous dampers, 52 tuned masses
  11. 6:07 It had already happened in Japan
  12. 6:41 A frequency question for a stability problem
  13. 7:25 What changed in the codes
  14. 7:58 Three episodes, one shape

Transcript

Open on Saturday, shut on Monday 0:00

On the tenth of June, two thousand, London opened a footbridge across the Thames. Ninety thousand people crossed it that day, up to two thousand of them standing on the deck at once, and the deck began to move sideways. Not a tremble. Seventy millimetres, back and forth, enough that people stopped walking and took hold of the handrail. The bridge was closed two days later. It stayed shut for twenty months.

You have almost certainly heard the explanation. The crowd fell into step -- with each other, and with the bridge -- and the pushes added up. It is in textbooks, in documentaries, and in the popular science press as the classic example of synchronisation. The researchers who study this phenomenon now say that the stepping in step was a result of the instability, not the cause of it. That is this episode.

A bridge built to look like it is not there 0:56

Start with the object, because it is unusual. Three hundred and twenty-five metres of walkway across the Thames in three spans, the longest of them one hundred and forty-four metres. The deck is four metres wide. It is a suspension bridge with the cables pulled down almost flat and carried outside the deck rather than above it, so that nothing rises above the handrail. Eight cables, tensioned to two thousand tonnes. It was built to look like it is not there.

Opening day, in measured quantities 1:27

Here is what the opening day was, in measured quantities rather than adjectives. Crowd density on the deck reached between one point three and one point five people per square metre. The lateral acceleration measured on the structure reached between one point eight six and two point four five metres per second squared. That is roughly a quarter of gravity, sideways, under your feet. At that level you do not stroll. You widen your stance and you hold on to something.

The explanation everybody repeats 1:59

The classic explanation goes like this. Walking is not a symmetrical act. Every time you put a foot down you push sideways, left, then right, so your body produces a lateral force at about one Hertz -- half your step rate. Normally those pushes are scattered across a crowd and cancel out. But if the deck itself begins to move at that frequency, it is easier to walk in time with it than to fight it, so people fall into step, the pushes line up, and the motion grows. A crowd of metronomes on a shared board.

Arup puts people on the bridge 2:34

The engineers did not argue about it. Arup, who designed the bridge, closed it and ran tests on the structure itself, walking groups of volunteers across the deck and making the groups larger and larger while instruments recorded what it did. A threshold appeared, and it appeared sharply. Below a certain number of people on a span, nothing happened at all. Add a few more and the motion did not increase a little. It started to grow on its own. On one span that number was around one hundred and sixty-six people.

The force followed the speed 3:09

Now the part that puts this bridge on this channel. When Arup looked at the force the walkers were putting into the deck, it did not follow the deck's position. It followed the deck's speed. A force proportional to velocity has a name in engineering, and the name is damping. Damping is the thing that takes energy out of a moving structure and brings it back to rest. What the crowd was supplying was damping with the sign reversed. The bridge was not being pushed. It was having its brakes taken off.

The critical number of pedestrians 3:43

That turns the threshold into arithmetic. Arup published a formula for the critical number of pedestrians: eight pi, times the frequency, times the modal mass, times the damping ratio, divided by a measured coefficient for how much negative damping a single walker supplies. Every person on the deck subtracts a little damping. Once the crowd has subtracted more than the structure had to begin with, the total is negative, and any disturbance at all grows instead of dying away.

2021: synchronisation is not required 4:15

Which brings us to the sentence everybody repeats, and to two thousand and twenty-one. Seven researchers went back through the observations, the experiments and the models, and published the result in Nature Communications. Their argument is that synchronisation is not required for any of this. Uncorrelated pedestrians -- people not in step with each other or with the bridge -- still produce negative damping on average, because their individual contributions do not cancel. On the Millennium Bridge itself the evidence for synchronised footfall covers at most twenty per cent of the walkers, and it shows up after the deck is already moving.

Why the distinction is not semantics 4:54

That is not a quarrel about words, and here is why. If the cause is synchronisation, the defence is to keep the bridge away from the frequency a crowd can lock on to. Put the lateral modes outside the range of a walking pace and you are safe. If the cause is average negative damping, that defence does not hold, because the effect is not tied to matching a frequency. It appears across a wide band. The only thing that reliably beats negative damping is positive damping, and positive damping has to be put there on purpose.

37 viscous dampers, 52 tuned masses 5:29

Which is exactly what fixed the bridge. Not stiffening, not retuning. Damping, added as hardware. Thirty-seven viscous dampers went in -- seventeen in chevron pairs under the deck, four vertical, sixteen at the piers -- along with fifty-two tuned mass dampers. The damping of the structure went from about half of one per cent of critical to around twenty per cent. Five million pounds, against eighteen point two million for the bridge itself. It reopened in February, two thousand and two, and it has not done it since.

It had already happened in Japan 6:07

Now the uncomfortable part. None of this was new in the year two thousand. Seven years earlier a Japanese team led by Yozo Fujino had published measurements from a crowded footbridge at Toda Park. The deck swayed sideways under a dense crowd, and about twenty per cent of the pedestrians ended up walking in time with it. That was in a mainstream structural dynamics journal, in nineteen ninety-three. The phenomenon was not unknown. It was in the literature, and it was not in the design code.

A frequency question for a stability problem 6:41

So the root cause is not the crowd, and it is not the flattened cables either. It is the question the check was asking. Footbridge design asked whether anything could drive the structure at its own natural frequency, and told you to keep those frequencies away from the rate a person walks at. That is a resonance question. Everything in the code was built around the idea that a structure gets into trouble when something outside it pushes in rhythm. This is not resonance. This is an instability: the structure loses more damping than it owns, and after that the frequency does not matter. You cannot find a stability problem on a list of frequencies. Nobody was looking for one.

What changed in the codes 7:25

What changed afterwards is the shape of the check. Lateral behaviour under crowds went into the design guidance: the French Setra guide in two thousand and six, and the British national annex to the Eurocode. With it came a rule that had not existed. A footbridge with a lateral mode below roughly one point three Hertz needs a stability calculation, not just a frequency table. Damping stopped being a number you hope the finished structure turns out to have, and became a quantity somebody is responsible for providing.

Three episodes, one shape 7:58

Three episodes now, and the same shape underneath all three. A deflection theory that was right for every bridge it had been tested on. A code checklist that was complete for every tower built before it. A vibration criterion that was correct for every way a footbridge had been known to misbehave. None of the three was wrong. All three were finite, and none of them said so on the page. Next time we leave the ground. In nineteen sixty-five the allowable voltage on a heater inside an Apollo oxygen tank was raised from twenty-eight volts to sixty-five, and one component in that circuit was never changed to match. Five years later, on the launch pad, it welded itself shut. Nobody found out until the tank was three hundred thousand kilometres from Earth.

Description and sources

Ninety thousand people crossed the new Thames footbridge on opening day, up to 2,000 of them on the deck at once, and the deck moved sideways hard enough that people stopped walking and grabbed the handrail. Arup closed the bridge and tested it with walking groups of increasing size, and found a sharp threshold — around 166 people on a span — above which the motion no longer grew in proportion to the crowd but grew on its own.

The force they measured tracked the deck's speed rather than its displacement. A force proportional to velocity is damping, and this one had the wrong sign: every pedestrian subtracted a little of the damping the structure had. Past the critical number the total went negative and any disturbance grew. That is the same mechanism as the Tacoma Narrows deck in 1940, and it is not resonance.

In 2021 seven researchers argued in Nature Communications that the synchronisation everyone repeats is not required at all — uncorrelated pedestrians produce negative damping on average, and the coherent stepping is a consequence of the instability rather than its cause. The root cause sits further back: footbridge design asked a frequency question, which is a resonance question, and this is a stability problem. A stability problem does not appear on a list of frequencies. The phenomenon had already been measured and published on a Japanese footbridge seven years earlier.

PRINT-READY, FROM THIS CHANNEL

The Failure Atlas, Vol. 01 — Tacoma Narrows · Citicorp Center · Millennium Bridge · Apollo 13 · the 2003 blackout · Hyatt Regency

https://therepository.gumroad.com/l/failure-atlas

PRIMARY SOURCES

The 18 technical plates in this video are illustrations generated for the channel by a diffusion image model, styled to match its cyanotype identity. They are diagrams of the system, not photographs of the hardware, and no person is depicted in any of them.

Root Cause investigates why engineered systems fail, using the official investigation reports and the primary technical literature. Sources for this episode are linked above.

The technical drawings in this video are cyanotype-style illustrations produced for the channel. They are diagrams, not photographs of the real hardware. The charts and dimensioned comparisons are drawn from the figures in the sources listed above.

#engineering #failureanalysis #rootcause #bridges #london