Air France 4590 lifted off from runway 26 Right at Charles de Gaulle at 16:43 on Tuesday 25 July 2000, already on fire, and hit a hotel at Gonesse about seventy seconds later. Everyone aboard died — 100 passengers and nine crew — along with four people on the ground.

The aircraft that had never killed a passenger in twenty-four years of service killed 113 people in under two minutes.

Almost everyone knows the outline: a strip of metal on the runway, a burst tyre, a punctured fuel tank. Two of those three are right. The third is the part the official investigation went to considerable trouble to correct, and it is still repeated wrongly almost everywhere — including in a French court.

Five minutes earlier

Continental Airlines flight 55, a DC-10 registered N13067, took off from the same runway five minutes ahead of the Concorde. It left behind a strip of metal from the thrust-reverser cowl door of its number three engine.

The BEA measured it: 435 mm long, 29 to 34 mm wide, about 1.4 mm thick, with twelve drill holes at random spacing, some off centre. It was a titanium alloy — roughly 90% titanium, 7% aluminium, 2% vanadium.

It should not have been titanium. The manufacturer told the BEA that a workshop-made replacement strip had to be made of stainless steel to comply. The strip had been fitted at Houston on 9 July 2000, and the BEA’s finding is unambiguous: it “had neither been manufactured nor installed in accordance with the procedures as defined by the manufacturer”.

Only one other aircraft, an Air France 747, used the runway between the DC-10 and the Concorde.

16:42:31

The takeoff roll began. V1 was called at 16:43:03.7. Somewhere in the following ten seconds, tyre number 2 — the right front tyre on the left main gear — ran over the strip at high speed and was destroyed. A piece weighing 4.5 kg, roughly a metre by a third of a metre, was later recovered from the runway. More than 30% of the tyre was gone.

Nine seconds after V1, the first officer said “watch out”. Four-tenths of a second after that, the controller transmitted.

The aircraft lifted off at 16:43:21.9. It was already burning.

An airline captain’s technical reconstruction of the accident sequence, built on the BEA report.

What actually opened the fuel tank

This is the part that almost every retelling gets wrong, and it matters, because the wrong version makes the accident sound like bad luck and the right version makes it sound like physics.

The popular account is that a chunk of tyre was thrown upward and punched a hole in the wing tank. The BEA examined the piece of tank recovered from the runway and ruled that out.

The recovered panel, about 32 by 32 centimetres, bore no impact marks at all. It had been ripped away from the inside towards the outside.

Tank 5 was about 94% full. Fuel does not compress. When the 4.5 kg piece of tyre struck the underside of the wing, the energy went into the fuel, the fuel moved, and the skin failed at a weak point somewhere else. The BEA’s own summary of the mechanism calls it “a complex process of transmission of the energy produced by the impact of a piece of tyre at another point on the tank”.

It also modelled the hydrodynamic pressure surge that everyone now quotes as the explanation, found that it was real, and found that it was not enough on its own: “the loads transmitted to the structure did not lead to a rupture in the area affected by the pressure surge”. In the probable causes it appears only as a possible contributory effect.

The report is candid about the limits of what it could establish. It describes a failure mechanism “that had never been seen on civil aircraft before the accident and about which it is difficult to determine the precise process”. The UK accredited representative dissented from the BEA’s preferred scenario, arguing that the two mechanisms could not be separated on the available evidence and that the real failure may have been a combination. The BEA published that dissent and declined to change the text.

The ignition source was never settled either. An electric arc in the landing gear bay, or contact with hot parts of the engine: the specialists working on the investigation could not agree which was more probable. What is certain is that the fire came first and the engine surges came after.

Seventy seconds

The crew were Captain Christian Marty, First Officer Jean Marcot and Flight Engineer Gilles Jardinaud. The BEA names none of them — its transcript identifies speakers only by seat — but two first names are audible on the tape.

Engines 1 and 2 surged almost together. The flight engineer called “failure engine two”, then “shut down engine two”; the captain called for the engine fire procedure. The first officer was already calling airspeed. The captain called for the gear up, and it would not come up.

That last point is in the probable causes as well. With the gear down, the flame stayed attached and stabilised under the wing for the whole flight.

At 16:44:14.6 the first officer said “Le Bourget, Le Bourget” — the small airport off the nose. The captain’s replies, marked in the transcript as uncertain words, were “too late” and “no time, no”. The first officer told the controller they were trying for Le Bourget anyway. The ground proximity warning had already sounded three times.

The recording ends at 16:44:31.6. The last thing on it from the captain is not words. The BEA’s annotation reads, in translation, “noises of effort”.

The report’s judgement on whether any of it could have gone differently is one sentence, and it is the sentence that should end every argument about crew performance on this flight.

The things that did not cause it

Three details get pulled into the story and all three were examined and dismissed by the investigation.

The weight. The aircraft was over its maximum structural takeoff weight, by about a tonne once the fuel not burned during taxi is counted. The BEA’s finding: “Any effect on takeoff performance from this excess weight was negligible.”

The centre of gravity. It was roughly 54% at takeoff, in accordance with the Flight Manual recommendation.

The missing spacer. A spacer in the left main landing gear bogie had not been refitted when the bogie was replaced a week earlier. The BEA called the omission a grave error and said plainly that it did not contribute to the accident. A former Concorde captain and flight engineer later put a sixty-page counter-argument to the investigating judge, contending that drag from the missing spacer caused the aircraft to rotate further down the runway than it should have — over the strip rather than before it. The BEA disputed it, saying the acceleration was normal. It is a live dissent, not an official finding.

There is also a tailwind in the record — the controller passed 090 degrees at 8 knots — which the BEA noted the crew did not comment on, and which it found “surprising”.

Pontoise, then Versailles

The criminal trial opened at the tribunal correctionnel de Pontoise on 2 February 2010 — not, as is often written, in Paris. In the dock: Continental Airlines as a company, the mechanic John Taylor who made and fitted the strip, his supervisor Stanley Ford, and three men from the French side of the Concorde programme — Henri Perrier, Jacques Hérubel and Claude Frantzen.

On 6 December 2010 the court convicted Continental and Taylor of involuntary homicide and acquitted the other four. Continental was fined 200,000 euros, ordered to pay Air France a million, and made to carry 70% of the compensation liability. Taylor received fifteen months suspended.

Continental’s response was not diplomatic.

The case went on appeal to the Cour d’appel de Versailles, which opened on 8 March 2012. Henri Perrier died in May of that year, aged 83. On 29 November 2012 the appeal court overturned both criminal convictions and quashed the fine — while leaving the civil liability, and the million euros to Air France, standing.

The presiding judge’s preliminary declaration is the most interesting document to come out of the whole legal process, and almost nobody has read it. It is not really about a metal strip at all.

Concorde’s fuel tanks had been known to be vulnerable since a 1979 incident at Washington. The fleet was so small and flew so little that a failure rate which would have been intolerable on an A320 took twenty-four years to produce a catastrophe.

Three more years

Air France grounded its Concordes the same day. British Airways did not: it cancelled that evening, resumed the next morning, and kept flying until 15 August 2000, stopping only when it learned what was coming.

On 16 August 2000 the French DGAC and the UK CAA suspended the individual certificates of airworthiness of every Concorde, on a joint recommendation from the BEA and the AAIB. The type certificate was never withdrawn — a distinction that gets lost, and that mattered enormously to whether the aircraft could come back.

It came back. A joint airworthiness directive on 5 September 2001 mandated Kevlar liners in the bottoms of six fuel tanks, Michelin Near Zero Growth radial tyres on all eight main wheels, and armoured electrical wiring in the undercarriage bay.

Both airlines returned to service on 7 November 2001. British Airways had flown a dress rehearsal with its own staff aboard on 11 September 2001; the aircraft landed back at Heathrow that lunchtime, into a world that had changed while it was airborne.

Eighteen months later it was over. On 10 April 2003 both airlines announced retirement on the same day.

The public line was a joint commercial decision: premium travel had collapsed, maintenance costs were climbing, Airbus said the support burden was growing fast with age. British Airways was filling just under half its Concorde seats; Air France was filling about a fifth.

British Airways’ chairman told The Times something rather different three weeks later.

Air France’s last commercial service, AF001 from JFK, landed at Charles de Gaulle on 31 May 2003. British Airways ran until 24 October 2003, when three Concordes landed at Heathrow within four minutes of each other. The very last Concorde flight of any kind was on 26 November 2003, Heathrow to Filton, watched by a crowd of more than twenty thousand.

The last passenger service into Heathrow, 24 October 2003.

What to be careful about

Six things repeated constantly about this accident are wrong or overstated.

The tank was not punctured. The BEA explicitly excluded direct puncture as the cause of the large rupture. The panel was pushed out from inside by displaced fuel.

The shockwave did not burst the tank on its own. It was modelled, found real, and found insufficient. It appears in the probable causes as a possible contributory effect.

The aircraft was not dangerously overweight. About a tonne over, with negligible effect on takeoff performance.

The missing spacer did not cause it — officially. There is a serious dissenting argument, but it is a dissent.

It was a court at Pontoise, not Paris. And Continental was cleared criminally in 2012, not civilly: the million euros to Air France survived the appeal.

The 2010 court said the strip should have been aluminium. The BEA says the manual specified stainless steel. Both statements get printed as if they were the same finding. They are not.

And one that is true, and worth sitting with: the BEA concluded the aircraft was lost from the moment the tank opened, whatever the engines did afterwards.

Sources: BEA, Rapport final sur l’accident survenu le 25 juillet 2000 au lieu-dit La Patte d’Oie de Gonesse (f-sc000725a), English translation, 16 January 2002, including Appendix 2 (CVR transcript); UK Civil Aviation Authority press release, 5 September 2001; Air France and British Airways retirement announcements, 10 April 2003; BBC News, 2 February and 6 December 2010 and 29 November 2012; Que Choisir, 16 December 2012; Ben Webster, The Times, 1 May 2003; FAA Lessons Learned Library.

A full-length documentary treatment of the investigation.