They Searched 120,000 Square Kilometres For MH370. It Is 1,100km Away.

PERTH, Australia — Two hundred and thirty-nine people were aboard Malaysia Airlines flight MH370 when it stopped talking to the ground on 8 March 2014. Their families have been waiting twelve years. This newspaper has spent nine months on the signal record, and we think the search looked in the wrong place — not by a little, and not by accident.

The aircraft kept one line open after everything else went quiet.

A satellite terminal aboard the aircraft continued to answer automated status requests from an Inmarsat ground station for close to seven hours after the transponder stopped. Seven of those exchanges were logged. They carried no position. They carried something that can be turned into a position: a timing value, and a frequency value describing how much the aircraft's motion had shifted the signal.

The timing value draws an arc. The frequency value says where on the arc the aircraft could plausibly have been, given how a Boeing 777 actually flies.

The published reconstruction put the seventh arc in the southern Indian Ocean and sent the ships there. Between 2014 and 2018, four vessels covered roughly 120,000 square kilometres of seabed at depths reaching 6,000 metres. They found nothing belonging to the aircraft.

They were looking at the right arc. We do not think they were looking at the right part of it.

The offset that was corrected twice

The burst-frequency offset is not a raw measurement. Before it means anything, it has to be corrected for the satellite's own drift, for the ground station's compensation, and for the temperature behaviour of the aircraft's oscillator — a component whose frequency wanders as it warms.

Aurelie Fenn-Massard spent eleven years working on the region's satellite links. She walked us through the correction chain over four calls, using her own notes rather than anything belonging to her former employer.

Her account is that the oscillator correction was applied against a warm-start profile. MH370's terminal had been powered continuously for hours before the logged exchanges began, and then, on her reading of the log, lost power and came back.

"A cold oscillator does not behave like a warm one," she said. "If you model a cold start as a warm one, you do not get noise. You get a bias, and the bias always points the same way."

She put the size of that bias at between 4 and 7 hertz. Applied across the final three exchanges, it moves the most probable terminus northeast along the arc by between 900 and 1,300 kilometres.

The centre of our reconstruction sits about 1,100 kilometres northeast of the nearest edge of the area that was actually searched.

What washed up, and where

Twenty-seven pieces of debris attributed to the aircraft have come ashore since 2015 — on Réunion, Mozambique, Tanzania, South Africa and Madagascar. A flaperon was the first, in July 2015.

Drift modelling has always sat awkwardly with the official terminus. Ruben Okonjo-Halstead has run reverse-drift studies on the western Indian Ocean for two decades.

"You can get debris from the searched box to Réunion," he told us. "You have to work at it. The arrival times come out late and the spread comes out narrow, and the real arrivals were early and wide."

He ran our coordinates through the same model. The arrival window moved to within five weeks of the observed dates, and the spread widened to match.

"I want to be careful," he said. "Drift modelling cannot find an aircraft. It can tell you when a starting point is uncomfortable. The published one has always been uncomfortable. Yours is less so. That is all I will say."

That is all we will claim for it.

What we could not establish

We do not know why the correction was applied the way it was. We asked. We have been asking for six months.

We put the oscillator question to the satellite operator, to the Australian agency that ran the search, and to the Malaysian ministry that holds the investigation file. The operator declined to discuss the correction chain. The Australian agency referred us to its published reports, which do not address a cold-start case. The Malaysian ministry did not reply to four letters.

We also cannot say what happened aboard the aircraft, and we are not going to guess. The signal record describes a machine. It says nothing about the people inside it, and the families have had twelve years of other people's guesses.

Marisol Ibbotson-Vare's brother was in seat 34K. She asked us for one thing before we printed any of this.

"Say how many," she said. "Everyone writes about the plane. Two hundred and thirty-nine people got on it. Start there, or don't write it."

We have started there. The seabed northeast of the searched box has never been surveyed. It is deep, it is remote, and it would cost money that nobody has volunteered.

It is also, on the arithmetic above, where the aircraft most probably is.