麻豆传媒

A potential breakthrough in one of the longest-running projects in physics showed how science really works, says sociologist Harry Collins

THE two huge interferometers belonging to one of the world鈥檚 biggest physics experiments, the Laser Interferometer Gravitational-Wave Observatory (LIGO), experience a simultaneous burst of energy. The interferometers are separated by around 3000 kilometres, one in Louisiana and one in Washington state, and such interesting coincident pulses are just what the scientists are looking for鈥�

That was in September 2007. Chance aside, a coincidence across that distance had to be caused by something interesting, so it began to look as if this was the nearest thing to a believable gravitational wave anyone had seen in 50 years of searching.

The next 18 months were exciting. Behind closed doors the scientists worked and argued over the 鈥淓quinox Event鈥�. Was it a signal? Could it, after all, have been a chance coincidence caused by noise? Then again, was it a 鈥渂lind injection鈥� 鈥� a fake signal introduced by members of the team sworn to secrecy? The idea being to keep the scientists on their toes and to remind them not to be so sceptical they lose the ability to see anything after 50 years of collective suspicion about every last blip.

鈥淏ehind closed doors the scientists worked and argued over the Event鈥�

In the early 1970s, my sociology PhD had taken me to the US on a road trip. Among the four areas I set out to explore, one I learned about through a 麻豆传媒 article. I have now spent nearly 40 years investigating gravitational wave detection, and have recently been admitted to the meetings and email discussions that are normally the preserve of insiders alone. I have been recording the debates and photographing key moments. It鈥檚 a unique position for a sociologist, professionally and personally, and some of the scientists are now my friends. The full story of the investigation of the Equinox Event and its tense and dramatic conclusion are set out in my book Gravity鈥檚 Ghost.

Einstein鈥檚 theory of relativity predicts that movements of massive objects, so long as they are not exactly symmetrical, will generate disturbances in space-time. These spread out like waves. Strong waves might be generated by the final moments of a binary star system as the two bodies spiral into each other, for example, or by the explosion of stars.

The 1993 physics Nobel prize was awarded to Russell Hulse and Joseph Taylor, both of Princeton University, for their work on the decaying orbit of a binary star system on the journey that would eventually lead to its collapse. They concluded that the decay was consistent with the emission of gravitational waves. But the ultimate goal is to detect the waves directly. When that is achieved we will have a new form of astronomy that can see into the heart of quaking black holes.

The trouble is that the strength of gravitational waves decreases with distance from the source. That means that by the time even the strongest waves get to Earth, they will be very weak, so the hunters need very sensitive detectors. The LIGO detectors are state of the art, but even these massive machines have little chance of seeing the effect.

To claim a detection, the American interferometers need to establish with high statistical significance that there have been coincidental changes of 1000th of the diameter of a proton in the 4-kilometre length of the interferometers鈥� arms. The search for gravitational waves involves extracting vestigial signals from the noise and working out that they were very unlikely to have been due to chance. This requires enormous care and painstaking protocols to avoid any possible bias.

As the story of the unravelling of the Equinox Event shows, many of the standards set in advance could not be met. For example, one rule is to 鈥渇reeze鈥� the statistical protocol before analysing the data and then never change anything in case this is seen as biasing the statistics. But this rule has some crazy implications, such as not being able to cross-check a result with other kinds of astronomers the moment something that could be important is seen: as soon as you know you might have seen something, the subsequent analysis cannot be so neutral.

Moreover, statistical analysis in this, as in all sciences, depends on all kinds of unknowns and imponderables: how many interpretations of the statistics have been made by others? Was a particular outcome imagined before or after the event? And if weak results and their attendant hypotheses are not published (something the scientists tend to hate doing), how can we be sure that hypotheses are not being continually altered to fit the facts 鈥� 鈥渄rawing the bulls-eye after the arrow has hit the target鈥�, as one of the scientists put it. Statistical tests are like used cars: they may look shiny, but you need to know the number of previous owners and how the cars have been driven. In statistics as with cars, it is usually impossible to know for sure.

Nearly every decision made in the course of working out the Equinox Event turned out to have been affected by the field鈥檚 history. The years of failed claims had been experienced as intensely embarrassing by the scientists 鈥� and used as ammunition by their critics. The net effect was an endemic cautiousness: this has pushed young scientists to bide their time and they are waiting for the generation of more powerful detectors, now being built, to come online. Some of the older scientists, with less working time ahead of them, are less patient.

In spite of all these problems, the unfolding of the Equinox Event represents science at its best, albeit a fallible best. The trouble is that scientific endeavour, especially in exact sciences like physics, is still masked by a mythology of perfection, something the scientists involved find hard to give up.

That is why so little is known about the Event: the scientists would say nothing in public even about such an interesting stretch of data unless they were certain they had found a gravitational wave. Obviously, the Equinox Event never reached the standard for announcement as a discovery or you would already know about it. Nevertheless, since the dust has settled, I have been freed to give the story the richer telling it deserves, to try to recapture the excitement of those months, and to try to draw out the lessons that can be learned from it.

I found researchers laudably aspiring to a standard of perfection that just could not be attained. But the integrity of the enterprise can still provide lessons on how to go about making technological decisions in conditions of uncertainty. It shows that something better than cynical realpolitik can and should always be our goal.

And if the rest of us are to be able to draw these lessons we need to be allowed to see what really happens in science. We need more than the polished, hygienic accounts that appear in the final discovery papers. We need to learn from the struggle to make decisions in a good way as the intense and imponderable pressures of the moment bear upon them.