Experiments at the Large Hadron Collider - How are discoveries made?
With the LHC up and running, some might imagine physicists just waiting for a Higgs boson to pop up in one of the four experiments, before publishing a paper and moving on to solve science’s next Big Mystery. However this picture is very far from the reality of experimental particle physics today, where results are based on statistics, statistics and yet more statistics.
Simulated Higgs decaying into four muons in the ATLAS detector.
This is because new particles are hardly ever detected directly. Depending on the design of the experiment, there can be several centimetres between the collision point and the first layer of detector electronics. This is a vast distance on the scale of a subatomic particle. Particles created in the collisions interact and indeed decay before ever reaching the detector. If created in an LHC collision, the Higgs boson, for example, is expected to last for a mere trillionth of a trillionth of a second*, before decaying into other particles – not long enough to move anywhere perceptible, let alone reach the detector. Moreover its decay products may themselves also decay before being detected.
It is only through a complete analysis of the tangle of particles created in collisions that physicists can begin to suppose the existence of something new. Over time, with repeated observations of the same effect, suppositions become more certain. The larger the data set, the more confidence we can have about the results. However, new particles are not created in every collision. Far from it! Finding a Higgs will be like finding a needle in a million haystacks.
Making a discovery often consists of collecting a lot of data and combining them in a plot (one of the most common plots is production rate against mass) before carefully subtracting the contributions expected from known processes. This is because known particles can often produce similar decay products, or signatures in the detector, to new particles. Once all known processes are accounted for, if a signal remains, it might be down to something new.
Often you will hear a sigma value attached to a particular announcement. This gives a measure of how certain physicists are that the result is real and not just a statistical fluke. A two-sigma result has a 2.3% chance that it is not true. A three-sigma result has a 0.15% chance that it is not true. Beyond the three-sigma level, physicists start sitting up and taking notice, but a result can’t yet be termed a new discovery. For that to happen, a five-sigma certainty is needed, or a mere one in 3.4 million chance that a sighting is due to chance.
And although a three-sigma result may already sound pretty certain to you, don’t forget physicists are making huge numbers of different plots. If you plot 1000 different distributions, the chances become about one in a thousand (0.1%) that you'll see something odd in one of them, without it being due to a new particle.
Luminosity is key - the more protons in the LHC and the higher the collision rate, the more data are collected and analysed by the experiments and the lower the error. But this only works up to a point. The sigma value alone does not give the whole story. The contributions from known particles that are subtracted from the plots are not always known exactly. This is a potential source of systematic error that is not reflected in the sigma value.
In addition, the choices made by the physicists running the experiment, such as the criteria for which data to store and analyse, can contribute to errors in the analysis. These choices are based on our current understanding of physics, which gives pointers to what we expect to find, but they are not infallible. There can also be effects due to the way a particular experiment is designed.
Hence the importance of comparisons with results from different teams. CERN’s communication protocol ensures that before the lab makes an announcement of a new discovery, the leaders of the other LHC experiments have also had a chance to compare results to their own measurements and give their reactions.
This doesn’t mean it’s wrong to get excited about the first indications of new physics. It is just important to put them into context. They should be taken with a pinch of salt until they have been both tested over time and subjected to peer review. With the LHC performing so well, there is great excitement all over CERN and huge anticipation of the results that are on their way. Roll on the summer conference season and its abundance of new peer-reviewed papers.
* 10-24 s
by Emma Sanders