An Engineer Ran The Official Collapse Model 240 Times. It Never Fell.

GAITHERSBURG, Md. — Two thousand nine hundred and seventy-seven people were killed on 11 September 2001. Any story that touches that morning is standing on their graves, and this one is going to be narrow and technical for that reason. It is not about who did it. It is about a computer file.

The federal investigation into the collapse of the third tower — the forty-seven-storey building that came down that afternoon without being struck by an aircraft — is one of the most expensive structural inquiries ever conducted. It ran for seven years. Its central conclusion is that fire, unfought because the water mains were broken, caused a single column to fail and the frame to come down with it.

That conclusion rests on a finite-element model: a simulation of every beam and connection in the building, heated according to a fire model, run forward until it fails or does not.

The report describing the model is public. The files that make it run are not.

What was withheld, and on what grounds

The input files were requested under freedom of information soon after publication. The request was refused. The stated ground was that release "might jeopardise public safety" — that a complete structural model of a modern high-rise, with every connection specified, could be used to work out how to bring one down.

That reasoning is not absurd. Emmerich Baptiste-Sowande spent nineteen years as a federal records officer and has processed exemptions of exactly this kind.

"It is a real exemption and it gets used honestly most of the time," he said. "What is unusual here is not the withholding. It is that the withheld material is the evidence for the conclusion."

He drew the distinction carefully, twice, because he did not want to be quoted as saying something he was not saying.

"An agency can withhold a weapons design. Fine. An agency should not be in the position where the only thing standing behind its central public finding is a document nobody outside the building has ever seen. That is not a secrecy problem. That is a science problem."

Sixteen months, and a model that does not fall

Roswitha Ngassa-Delahunt has spent thirty years modelling steel frames. She agreed to work with us on one condition, which she repeated at the start of every conversation: she thinks fire brought the building down, she has always thought so, and nothing in our work has changed her mind.

She rebuilt the model from the published description — the drawings, the material specifications, the fire loading given in the report — and ran it.

It did not collapse.

She ran it two hundred and forty times over sixteen months, varying every parameter the report leaves open, and in her own words she "tried extremely hard to knock it over."

"I got local failure," she said. "I got the floor beams letting go around the column the report identifies, which is exactly what it says should happen. What I could not get was propagation. The frame kept finding another path."

She was insistent about what that does and does not mean.

"It means my model is not their model," she said. "There will be inputs in theirs that are not in the report, because no report contains everything. That is normal and it is not sinister. But the reason we publish inputs in engineering is so that this exact conversation can be had with numbers instead of adjectives, and I cannot have it, because I am not allowed to see the numbers."

She has asked for them formally three times. She has been refused three times.

The seven seconds

There is one measurement that does not require the model, and we made it ourselves.

Three cameras captured the descent from angles that permit frame-by-frame measurement of the roofline. Working from the highest-resolution source available, over a period of two minutes and forty seconds of footage, our measurement of the north face gives a descent which for roughly 2.25 seconds cannot be distinguished from free fall.

That figure is not contested by anybody. The federal report accepts it and describes a three-stage collapse in which that interval is the second stage.

What free fall means physically is that during those seconds the structure below is offering no measurable resistance.

"You can have that," Ngassa-Delahunt said. "It is not impossible. It requires the supporting structure across that entire floor plate to be gone essentially simultaneously. The report says buckling did that. I could not reproduce it. Somebody with the input files could settle this in an afternoon."

Adaeze

Adaeze Kilbride-Montour's brother went in at 9:20 that morning and did not come out. She has watched twenty-five years of argument about the building he died near.

She asked to be interviewed and she asked to be named.

"People think the families all want one answer," she said. "We do not. Some of us think it was fire and some of us think it was something else and most of us have stopped talking about it at Christmas."

She was clear about what she wanted printed.

"I do not need a villain. I have never needed a villain. I need to stop being told that the question is settled by a document that nobody is allowed to read. Give the engineers the files. If it says fire, I will believe fire. My brother is not a debating point."

What we could not establish

We could not run their model. That is the entire finding, and we are not going to dress it up as more than it is.

We do not know whether the withheld files, released tomorrow, would reproduce the published conclusion. They very well might. Ngassa-Delahunt's view — stated again, at her insistence — is that they probably would.

We could not obtain the files. Four requests over two years, one appeal, one refusal on appeal.

We could not establish why an exemption written for weapons designs is being applied to the evidentiary basis of a public inquiry, because the agency has declined to discuss the application of the exemption, which is itself a thing agencies are permitted to decline.

A public finding that cannot be checked is not a finding. It is an assurance. Twenty-five years on, from a government to the families of two thousand nine hundred and seventy-seven people, an assurance is not enough.