My blog is called the Large Idea Collider, which is referring to the Large Hadron Collider (LHC), the leading proton accelerator and collider in the world. I am a particle physics graduate student working at the ATLAS experiment at the LHC. This blog is now 6 months old, and I have yet to post something about particle physics. It’s like I have been avoiding the topic…
Well, no more! I can finally combine two of my passions together: particle physics and investigating wild claims. You may have seen these headlines around in the past week or so. CDF, one of the two experiments at the Tevatron, the second largest hadron collider in the world, has been announcing the potential discovery of a new particle. So what’s the story?
A group of physicists within the CDF collaboration was looking for a rare but known type of events. They were looking for the production of a pair of W bosons, or the production of a W with a Z boson. These particles are the carriers of the weak interaction, one of the four known basic forces of nature. They are of special interest because their masses, which are pretty high, would be a consequence of the existence of the Higgs boson. The Higgs boson would also be responsible for making the weak interaction different from the electromagnetic interaction, which is kind of important is life is ever to exist. As you may know, the central goal of particle physics today is to confirm or rule out the existence of the Higgs boson. If it turns out not to exist, physicists are going to have a lot of fun.
The W and Z bosons are highly unstable particles. They decay into something more stable way before we can see them directly. In order to see if the bosons were produced at the Tevatron, our group of physicists at CDF looked for collisions yielding a pair of high energy jets (essentially collimated sprays of particles made of quarks), an accompanying electron or muon, and a momentum imbalance, indicating the production of an invisible particle taking away a lot of energy.
When looking for this specific kind of event, what you really want to know is what other kinds of events may fake it. These other kinds of events are backgrounds. If you can estimate well enough how many background events you will get while trying to select signal events (the WW and WZ events), you’ve got yourself a particle physics analysis. You can never completely suppress the backgrounds, so you must know how much of it you will get in the end. For example, you estimate that with your super-clever signal selection, you should get 100 background events. When you pass your data through this selection, you count 150 events. You estimate from this that you have 50 signal events in your data.
Let’s get back to our CDF group. In their signal selection, they make use of a quantity designed to measure the mass of a hypothetical particle that decays in two jets. Let’s call it the dijet mass. Let’s have a look at the following graphics from some people over at the CMS experiment (a LHC experiment), which pretty much summarizes the whole situation.
Let’s focus on the part that doesn’t move first. The plot on the bottom shows the distribution of the dijet mass. What you see on the y-axis is the number of events featuring two jets with a dijet mass Mjj. The green distribution is the background expectation. The red distribution is the signal expectation. Both of these are derived from theory. The black dots are actual data, collected by the CDF detector. As you can see, the data matches the signal+background pretty closely, and there is a visible signal excess over the background. What the CDF group didn’t expect is the other excess right in the middle of the plot. This is what caused all the media fuss last week. An unexpected excess of data events with respect to the theoretical expectation.
Now, the question is how significant is that excess? Is it possible we measured the dijet mass wrong? Is it possible our signal selection is artificially creating that excess? Is it possible we forgot some known physics phenomenon that would generate an excess there? Can it be just a statistical fluke, like a bunch of 10 dices all yielding six after being thrown on the floor?
All these questions are very relevant and they must be investigated patiently if this analysis is to be turned into real science. Until these questions have been answered, it is useless to spend time trying to figure out what kind of new particle we may be seeing there. We may actually see only a flaw in our measurement. Let me present a short argument why I think this is the case.
In order to establish whether or not the unexpected bump is a statistical accident, we must know precisely how many events we would normally expect. The less precisely we know this number, the less seriously we can take any bump on top of it. This is called statistical significance. In this analysis, the WW and WZ signal they were looking for is found to have 5 sigma significance while the other bump has a 3 sigma significance. 5 sigma is the golden standard in particle physics, It means that there is approximately 5 chances in 10 million that what we see is a statistical fluke. For 3 sigma, it is 3 chances in 1000. 3 sigma is the lowest significance that can justify the publication of results, but it is never considered conclusive evidence. As you can see, we like to be damn sure.
Now, what happens if, say, our CDF group underestimated the uncertainty on the dijet mass quantity? This scenario is explored in the animated plot on the right. The dijet mass is varied by 5%. Getting the mass wrong by 5% is exceptionally easy, and there are plenty of factors that can give you such an uncertainty. We can see that when we play with the dijet mass this way, the unexpected bump comes and goes, while the expected signal bump stays. This shows why we don’t usually take a 3 sigma bump too seriously. A couple of oversights in the analysis can easily make it appear or vanish.
Since I haven’t looked in too much details at the analysis in question, I have no strong opinion as to what this oversight may be. All I know is that it is very hard to get everything right in an analysis, especially if your signal is rare and hard to catch. Even more especially if you weren’t looking for it. The analysis was optimized to find a signal in a narrow region around the expected WW and WZ signal bump, not to find evidence of new physics.
In conclusion, I don’t think the result is worth all the hype it has been getting. However, it is still troubling enough to warrant further investigation. Soon, the DZero experiment, the direct competitor to CDF, will release its own result on the same analysis, cross-checking the CDF measurement. Also, CDF will redo the same measurement with a larger dataset soon. If you never hear from this again, you will know why. If the bump is confirmed by further measurements, the fuss in the press will last a lot longer, and you can expect a lot of follow-up studies making the news too.