My first project as a CERN summer student was to assemble an electronic die. After a few hours of soldering and burnt fingers I produce this:
How does it work?
When you tap the die on a table a piezo sensor glued to the base ‘feels’ the impact and sends a small current through the printed circuit board and into a chip.
The chip converts the analog signal into a decimal value (for example a 1.24167 volt signal is converted to the number 1.24167) and reads the least significant decimal place to generate a random number. The LEDs then light up to show a number 1 to 6.
What’s the point?
Over the course of a week, and much to the annoyance of my office mates, I tap the die on my desk 1500 times, compile a data set of the numbers thrown and go about analysing whether the die is fair or biased.
What’s this got to do with CERN then?
Well statistical analysis is key to what CERN does and the discovery of the Higgs is a pertinent example.
In the Large Hadron Collider a Higgs boson is produced by approximately 1 in every 10 million particle collisions. The boson then decays in a fraction of a microsecond while the other collisions produce an array of other particles.
To make the hunt even more tricky, scientists didn’t know the exact energy level at which the Higgs would be found so they had to collide particles at a variety of energy levels.
So searching for the Higgs was like looking for a tiny needle in a massive haystack full of other needles where your needle exists for a minute fraction of a second and you don’t really know what it looks like.
Scientists therefore had to statistically analyse trillions of collisions to be sure that the small bump at 125GeV in the graph above was the signature of the Higgs and not just an unlikely random fluctuation.
Before announcing the discovery of a ‘Higgs-like particle’ in July 2012 scientists were 99.9999% sure they’d found their boson i.e. there was only a one in a million chance this was not the Higgs.
What’s that got to do with the die?
In figuring out whether the die was fair I produced a relatively large data sample then used statistical techniques to conclude with 95% confidence that any bias displayed wasn’t just a random fluke. So in a way the exercise was analogous to the search for the Higgs but on a much smaller scale.
Discover anything interesting?
After some intense number crunching, detailed analysis and complex modelling I concluded (drum roll please) – bet on the number 2.
It turned out my supervisor sabotaged my die. So the key lesson learned was – don’t dice with physicists.