Cell Fourteen Ran Hot For Nine Days. Its Thermocouple Was Four Millimetres Out.

DELFT, Netherlands — Four millimetres.

At the power levels this experiment works at, that is the distance between a discovery and a loose thermocouple, and it has been that distance since 1989.

This newspaper spent $71,000 establishing it.

What we bought

In January we commissioned a replication rather than writing about somebody else's.

Two laboratories. Forty Fleischmann–Pons-type cells, twenty each. Ninety days of continuous logging, no early looks, no discretion about when to stop.

The two groups had never corresponded and did not learn each other's identity until the ninetieth day. Loading ratios were sealed at filling and opened at teardown, so that nobody was reading a thermometer while knowing what the cell was supposed to do.

The protocol, the calibration schedule and the stopping rules went up before the first cell was filled. That was the point. Anybody minded to accuse us of moving a goalpost could go and read where we had planted them.

What thirty-nine cells did

Nothing.

Thirty-nine of the forty sat inside calorimetric error for the whole ninety days. Not near the edge of it. Inside it.

There is a residual and it is real and it is in every cell: roughly 0.3 watts at the cell head, which is deuterium and oxygen recombining on the catalyst that is there to make them recombine. Laboratory One puts it at 0.31 W. Laboratory Two puts it at 0.26 W. Neither can close that gap, and both say so in reports we have published entire.

Chemistry has known about that 0.3 watts since 1990. It is not energy from anywhere interesting. It is a cell doing what a sealed cell full of deuterium does.

What cell fourteen did

Cell fourteen ran 4.2 per cent hot for nine consecutive days.

We sat on it for three weeks, which was correct, and during those three weeks our science correspondent talked about it over dinner at a conference, which was not.

Then Dr. Ottilie Brandsma-Achebe opened the cell.

The drawing

The assembly drawing gives the position of the reference thermocouple as adjacent to heater block.

A technician read that as English. It was meant as a coordinate — the tap on the outer bath wall, 211 millimetres below the flange, which is where the other thirty-nine sheaths were found. In cell fourteen the sheath sat four millimetres off the heater.

The heater runs 6.4 K above the bath during a calibration pulse. A thermocouple four millimetres away from it is not measuring the bath. It is measuring the heater, and it is measuring it late, which is worse: the lag makes the calibration constant come out small, and a small calibration constant turns ordinary joule heating into excess power.

Corrected in the archived data — and the correction is arithmetic, not judgement — 4.2 per cent becomes 0.1 per cent. Inside the error, with the other thirty-nine.

The email we had already been sent

Here is the part that belongs to us.

On 3 February, before any money moved, the statistician this newspaper had retained wrote to the desk. The effect appears, on the field's own published incidence, in about one loaded cell in a thousand. Forty cells gives you roughly a four per cent chance of containing a single positive one.

"You are not funding a test," Dr. Mireille Sowande-Halloran wrote. "You are buying forty tickets in a thousand-ticket raffle and you have already written the sentence about how you didn't win."

We commissioned it anyway. Her email is published with this story, unedited, including the paragraph where she offers to design something we could afford.

Twenty-two years of line items

The other thing we obtained is duller and it took fourteen months.

A Navy laboratory at China Lake released the programme funding record for this work: fiscal 1989 to fiscal 2010, twenty-two consecutive years, no gap longer than a continuing resolution. Line items only — no findings, no narrative, no press release, ever.

The field was not starved of money. It was patiently financed, at modest scale, for two decades, and what it produced was not repeatable.

Professor Hana Videc-Ojo, who is in that record, objects to the previous sentence at length in her reply below, and her objection is better than our sentence.

What this desk takes from it

We ran the wrong experiment and we could have known it in February for the price of reading our own consultant's email.

The finding that survives is not about palladium. It is that at a third of a watt of signal, a four-millimetre assembly error and a historic discovery produce the same trace on the same chart, and always have — which means every claim in this field, in either direction, is a claim about somebody's build quality until a teardown says otherwise.

Thirty-nine cells did nothing, and that is worth almost nothing. Cell fourteen did something, and it was a thermocouple.