How to solve a problem like big G.

I've just returned from a Royal Society meeting at their new Chicheley Hall (cunningly designed to get scientists to actually talk to each other) titled: "The Newtonian constant of gravitation: a constant too difficult to measure?" There was so much in the meeting that was fascinating and amusing that it will take a few blogs to cover it, but my overall impression was that, although I have no bias either way whether G is constant, the experimenters are limiting themselves by their assumption that it is.

G is difficult to measure because gravity is such a weak force, and it stands alone theoretically so you can't (yet) get to it from other parts of physics. Most of the experiments to find G are done using a torsion balance. This is made up of two test masses joined like a dumb-bell and suspended from the crossbar's centre by a wire. Source masses are held horizontally to one side of the test masses to avoid confusion with the Earth's gravity which operates vertically. Gravity pulls the test masses sideways and the angle of twist of the wire is measured. Since the force needed to twist the wire a given angle is known, the experimenters can find the gravitational force due to the source masses and using Newton's F=GMm/r^2 they can infer G (knowing F, M, m and r, which requires very careful measurement, surveying and planning).

It is very clear that all the experimental groups believe that they have measured the correct value of G and their uncertainty in their work is low (within 50 parts per million, ppm). However, their different values for G differ by 480 ppm (ten times the uncertainties)! No one knows what is causing these significant differences. The experimenters suspect overlooked mundane effects, others, eg: Prof Gibbons from Cambridge suggested at the meeting that perhaps G could vary in time. I wonder if inertial mass might vary in these experiments due to MiHsC, but the data to decide is not clear yet.

The experimenters themselves are a quality group: precise, tenacious, and also stressed, since they've spent the past decade measuring G, and there's been no closure since the true value is still unknown. It must be like training for 10 years for the Olympics, but at the end of the race it becomes clear that someone forgot to paint in the finish line and no-one knows who won. Dr Gundlach has given up on G and gone into biophysics and says he would never go back because of the sleepness nights (mind you, he smiles whenever he talks about his experiment, so he might). Dr Schlamminger (who seems to operate at twice the speed of everyone else) is young enough not to have run out of enthusiasm yet. Prof Clive Speake says that despite the huge import of this work he is finding it hard to engage the young in the classical-sounding physics of interacting balls. All the experimenters are keen to set up a new joint experiment that will solve all the (unknown) problems of the old ones and hope that the shared responsibility will mean less stress, but whatever new value for G they get is not going to agree with all the previous ones and may just add another data point to the menagerie.

In my view, a change of attitude is needed. One should look for patterns without presupposing a model. The obvious assumed model all the groups have is that G is constant. This means that they all take many measurements of G over, say, a month, and then average all these values to produce a precise G. What I think they should do, for old if possible, and new experiments is to publish time series of all the un-averaged data and all the environmental factors and experimental configurations in a data base online so others can look for patterns. The few plots I saw that did show the actual data (before averaging) showed large variations. What causes them? Are there any correlations? A compilation of old and new data like this would cost a tiny percentage of the cost of a brand new experiment, and curious scientists would then be able to search for patterns for free. Something like this was attempted by Dr Gillies at the meeting, but he only looked at variations in the source masses.

I don't know whether G is constant or not, but if it varies, or some other new physics is present they may never see it with their present averaging attitude. One should always test assumptions, and, in fact, that is the motto of the Royal Society: "Nullius in Verba: don't take anyone's word for it".

The webpage of the meeting is here:  http://royalsociety.org/events/2014/gravitation/