Saturday, 27 June 2009

Physics Fraud, and the Impossible Diamond

Physicists used to tell me was that physics was a special subject, because you never had to worry about the possibility of fraud. Their reasoning was that You Can't Fake Physics. If you make up an experimental result that isn't right, you're doomed to be found out when other people try the same experiment and can't replicate your result. It's a dumb thing to do, and no physicist would ever be stupid enough to try.

However, it might be more accurate to say that perhaps no sane physicist would try to fake a result that they believed to be wrong. Faking a correct result may be cheating, but doesn't carry the same risk. It's much more difficult to spot a fake result when it agrees with everyone else's results and with what everybody expects to happen.

We can sometimes spot a "false positive" when a theoretical prediction that is successfully verified later turns out to be wrong, or when an experimental technique later turns out to be impossible, or impossible to conduct to the claimed accuracy. When this happens in an experiment that contradicts current theory, we usually rip the person responsible to shreds, and accusations start flying of scientific fraud. When it happens in an experiment that agrees with current theory, we're usually more charitable, and tend to say that perhaps the experimenter was simply mistaken, or overcome with a little too much enthusiasm. There's such a large grey area for honest mistakes, or the unconscious selection of "good" data (or simple wishful thinking) that a certain amount of bad science probably slips under the radar without being spotted, and it's not often that we find a "bad" result supporting a "good" outcome that's really so profoundly impossible that people are forced to consider using the "f" word.

One candidate case happened in 1955.
Researchers had been wanting to create artificial diamonds since at least as far back as Nineteenth Century. When H.G. Wells published his short story "The Diamond Maker" in 1894, a number of researchers had already been trying approaches with varying degrees of optimism and claiming positive results, including James Ballantyne Hannay in 1880, and Nobel Prize-winner Henri Moissan (also in 1894). One of the wildest attempts to create artificial diamond was carried out by John Logie Baird, who briefly blacked out of part of Glasgow when he deliberately short-circuited an electricity substation's power terminals across a graphite rod embedded in reinforced concrete (the story goes that he couldn't work out how to get the thing open afterwards, and it ended up at the bottom of a river, unexamined).

The potential financial payoff for anyone able to create artificial diamonds on demand was obvious, and by the 1950's there had been more reported (but often disputed) successes, and competing researchers were trying desperately hard to be the first people to produce a proper, replicable, accepted process that definitely did produce diamonds. One team in particular figured that they were on the edge of actually achieving it. They had the theory right, they had the equipment right ... the only problem was that their pressure-vessel obstinately refused to cough up any diamonds.
It was desperately unfair. They'd done all the work correctly, and the experiment refused to come out the way it was supposed to. They needed a diamond to get further funding. From their perspective, they probably reckoned that they deserved a diamond. It was necessary for their future research. Science needed a diamond!

And a diamond dutifully appeared. They got new funding, bought new equipment and replicated the result, others managed the same thing, and everyone was happy.

Except that ... someone went back and checked the calibration on the original pressure reactor and found that its readings had been significantly "off". The pressure-vessel had been running at too low a pressure for diamond to form. With hindsight, their original artificial diamond seemed to have been a physical impossibility. So how did it get there?

Three of the four original team members put their names to a letter to Nature in 1993, explaining that subsequent spectral analysis of the "run 151" diamond years later had shown that it appeared to have the characteristics of a natural gemstone rather than those of an artificial rock. The experimenters had carried a small stock of natural diamonds for research purposes, and it seemed that one of those had somehow found its way into the pressure vessel during setup, and been "fortuitously" discovered after the experiment.

It's quite a nicely- and elegantly-written letter, but the authors must have been acutely aware that to most people, the idea that one might "accidentally" lose a real diamond inside an apparatus designed to create artificial diamond, in such a way that it could then be rediscovered and used to get further desperately-needed money ... if this happened in any other field, we'd tend to assume deliberate fraud.

Another thing that might surprise some outsiders is that although the announcement that the experiment had been a success was made in 1955, the retraction didn't happen until 1993, nearly forty years later. For Twentieth-Century experimental physics, this wasn't actually all that unusual – there seemed to be an unspoken "gentlemen's agreement" that if someone had claimed a "correct" result that they shouldn't have, that the community would hold off making too many pointed suggestions in print until some time after the person concerned was safely dead. This was probably a great way of avoiding public controversies, but it also meant that we never really got to the bottom of what had happened in many of these cases. If you weren't supposed to go public while someone was still alive, but you couldn't suggest fraud after they were dead (because it was unfair to level that sort of accusation at someone when they couldn't defend themselves), then it meant that anyone who did get up to no good had a decent chance of not being publicly outed, in print, ever. By the time a critical report could be written, the people with first-hand knowledge of what had really happened might have all died off.

By avoiding investigating these cases until after it was too late to reach a conclusion, the physics community probably did manage to achieve a nominal "no confirmed mainstream fraud" result. But that result was itself not especially honest.

Things are now looking up. Berkeley recently went public very quickly about problems with the work of two physicists (in two separate cases) who seemed to have been almost routinely fabricating data to get their "world-class" results (Victor Ninov and Jan Hendrik Schön), and there've now been a few more speedy "outings" of scientists caught misbehaving. So from now onwards, the more temptation-prone members of the physics community know that if they gain fame and fortune by faking data, universities and comissioning bodies won't necessarily hush the thing up for them.

But for research published before 2000 (or perhaps before ~2005) ... be more careful. A certain number of the "jewels" in physics history aren't quite what they appear to be.

No comments: