Wednesday, June 02, 2010

Fourth generation and scientific integrity

Nude Socialist has shown once again that the notion of scientific integrity means absolutely nothing for this magazine:
Matter: The next generation
Unfortunately, the fabrication of stories and distortion of the available evidence is being routinely applied not only to inferior scientific disciplines such as climate science.

In this case and many others, the most prestigious field of science, particle physics, became the victim. A writer called David Shiga decided that he would write an "exciting" story that there are probably four generations of quarks and leptons. So he has "improved" all the evidence to support such a story.

Let's look at the situation.

Quantum field theory and new generations

Classically, there's nothing wrong about including four (or more) generations of quarks and leptons instead of three. The anomalies cancel as long as the new generations contain particles with the same quantum numbers (charges and spins) as the known generations. The mathematical structure of the Standard Model may be easily extended to the case of a higher number of generations.

Also, in string theory, it is possible to found vacua that predict four generations of fermions. And some of them - I really mean four-generation models - look pretty natural. Well, there are also pretty natural three-generation models while the first Calabi-Yau manifold that was introduced into physics predicted 100 generations. Many other numbers are possible, too.

Quantum mechanically, too high a number of generations would also speed up the "running" of the coupling constant - the strengthening of the basic interactions as a function of an increasing energy that is caused by quantum mechanical effects (the buzzword is "renormalization group"). Even theoretically, you don't want to make the number of generations too high because your truncated effective field theory would break at too low energy scales.

Experimental data

More importantly, there exist loads of experimental data. None of them support the existence of the fourth generation of quarks and leptons. The situation is most obvious in the case of the neutrinos.

The Z-boson may decay to neutrino-antineutrino pairs. By measuring the rate of these "invisible" decay channels accurately 20 years ago or so, physicists could find out that the number of neutrino flavors was 3.01 +- 0.15 +- 0.05 (abstract). Clearly, the integers different from 3 are excluded at a 5-sigma or 6-sigma confidence level, to say the least.

Newest versions of the experiment have improved the estimate to 2.984 +- 0.082 which excludes four generations with a light neutrino at a 12-sigma confidence level.

It means that if there are additional neutrino flavors with the same quantum numbers, they have to be heavier than something like one half of the Z-boson mass or 50 GeV. If such a new neutrino were unstable, it would be ruled out by LEP II experiments; if it were stable, it would be ruled out by dark matter direct search experiments. At any rate, the hypothetical new generation would have to be "qualitatively different" because its neutrino would be very heavy. It still needs some neutrino because the anomalies wouldn't cancel without it. There are actually many other measurements that impose very strong and unnatural constraints on the hypothetical new generations. Fine. You may still think about such a new generation although the known facts make it very awkward.

The newest claims about the fourth generation are inspired by the recent paper
Search for heavy top t'Wq in lepton plus jets events in ∫Ldt = 4.6 fb-1
by J. Conway, D. Cox, R. Erbacher, W. Johnson, A. Ivanov, T. Schwarz, A. Lister, and - to a lesser extent - the rest of the CDF Collaboration.

The title sounds very technical but let's look at the abstract which should contain all "newspaper-worthy" main results of the research, shouldn't it?
We search for pair production of the heavy top (t') quarks pair decaying to Wq fi nal states using 4.6/fb data sample of lepton+jets collected using inclusive lepton and met+jets triggers.

We reconstruct the mass of the t' quark (M_{rec}) and perform a two-dimensional fi t of the observed (H_T ; M_{rec}) distribution to discriminate the new physics signal from Standard Model backgrounds. We exclude a Standard Model fourth-generation t' quark with mass below 335
GeV at 95% CL.
So the ultimate conclusion is that they could exclude the existence of a new quark at the 95% confidence level. The experiment shows it's not there. Of course, experiments are never "omnipotent", so they can rarely exclude a hypothesis for all values of its adjustable parameters. Because of the limited energy reach of the Tevatron, they could only falsify the theory at this confidence level if the new quark mass is assumed to be lower than 335 GeV.

Above this threshold, they can't say anything. The accuracy with which they can confirm the Standard Model drops because the number of relevant events goes down, too. The noise starts to dominate over the signals.

Now, you may see that the Nude Socialist text picks exactly the opposite qualitative conclusion. It wants to argue that this very paper shows evidence in favor of a fourth generation of quarks and leptons. That's a pretty manifest distortion, isn't it?

Does the paper contain positive evidence in favor of a new generation? Well, it contains some noise where you would expect noise. In particular, if you draw the graph of the cross section as a function of the energy, you may find a two-sigma deviation (and this deviation could be lowered by a model with an extra generation - but only if you adjusted it, too). Two-sigma bumps translate to the 95% confidence level.

So there's a 5% chance that such a bump emerges at a random, given place. The probability that such a bump appears anywhere, if you have at least dozens of places (in different graphs, plotting different quantities or correlations) where it can emerge, is close to 100%. It's not a real evidence for anything. It's just the "very overwhelming consensus" science that is supported by no nontrivial evidence.

Moreover, the same data or events that could be interpreted as a positive hint also falsify the explanation in terms of a fourth quark. The "fingerprints" simply don't agree.

John Conway - whom I know from Rutgers - the first listed author of the very preprint - added a comment to Tommaso Dorigo's blog (Dorigo has written a dozen of misleading articles on his blog, promoting the fourth generation) where he explained the fingerprints that point out to different explanations:
As one of the authors of the search, let me make a few comments... Firstly, I personally spent a very large portion of my research from last summer until March of this year working on getting the statistics of our results correct. With earlier, less sophisticated techniques, we were seeing that the excess had a much larger apparent excess. But it was only apparent - the true statistical significance of this excess has never been large, and still isn't.

Secondly, whatever the excess is, it is not a new 450 GeV quark. The excess events are all at very high total energy (HT) and spread out in apparent t' mass. This is not what we would expect to see, really, from a fourth generation quark. Also, the production rate for an "ordinary" fourth generation quark of such a mass is not large enough to account for the excess. And if the excess is not from t' we are at a loss to characterize just how significant the excess is, since we are observing it a posteriori.

Nevertheless the events are there, and show no signs of being some detector effect or misreconstruction. If there is a standard model background we have neglected we don't know what it is or we would have included it. We see 8 events in the highest HT bin where we expect about 2-3. No matter what it's not statistically significant, and as Tommaso points out we have seen such excesses before.

Is it interesting theoretically? Well, if some theorist out there thinks they have a model that we can test, particularly by looking at some other channel or variables we have not studied, we are all ears! There is nothing wrong with thinking about what our 8 events might be if they are not SM background or a t'. And we have had a few suggestions from some excellent, and quite real, theorists.
Fine. So there's a bump, it's very small. But even if you "inflated it" in size, its features won't agree with the predictions of a model with the fourth generation. And there exist better proposals what the measurement could mean if it we real. A classic attribution problem.

Misinterpretations and wishful thinking

But some people have already pre-decided, without any empirically-based rational reason, what the "right" story should be. And it's not just the inkspiller in Nude Socialist, David Shiga. It's only the same John Conway who gives an honest evaluation in the NS article:
Extraordinary claims require extraordinary evidence, and we definitely don't have that.
However, some of his colleagues have a very different idea what the purpose of science is. Stephen Martin says:
It's interesting enough that we'll be paying attention to future analyses and hoping. It would be very exciting if there was a [fourth-generation] quark.
I don't think it would be too exciting if there were a fourth-generation quark.

But more importantly, it's just not the job for particle physics experimenters to be "hoping". More precisely, they can be "hoping" in the silence of their private home but they shouldn't be "hyping" a speculation and they shouldn't encourage journalists to write dishonest stories about the research of their teams which is exactly what Stephen Martin did.

A related comment is that they shouldn't be paying attention to the conclusions of analyses that are "interesting enough" but to those that are "justified by sufficient scientific evidence". This is a classic confrontation between science, whose task is to search for the truth, and the sensationalist media, whose task is to search for sensations and ideologically (and financially) convenient conclusions whether they're true or not. Unfortunately, Mr Martin belongs to the latter camp.

Amarjit Soni, another researcher, is quoted as asking:
We've seen three generations, why not four?
In this context, this is a dishonest rhetorical question as well because it encourages the journalist - and the readers - to think that physics has no idea whether the number is three or four or something else. The question wants the reader to think that there's no good answer.

It's just not true. Physics ultimately does boil down to observations. So if there are only three generations that we may see, it means that there are three generations whether you like it (and understand it) or not.

Moreover, it's not just the "direct observation of particles" that implies that the number of generations is three and not higher. The invisible decays of Z-bosons measure the number of generations "indirectly" and the answer is three.

Also, there probably exist completely solid explanations of the number of generations based on string theory although the final explanation of this kind hasn't been found yet, due to the uncertainties about the vacuum selection. At any rate, it's no good to create the illusion that the number three is just a random guess.

CP-violation and fourth generation

The rest of the article brings another "argument" that is painted as an argument in favor of the fourth generation: the recent observational hints of CP-violation. It is very sensible to have serious doubts about the very reality of these signals, as the most recent results by CDF showed. Two-sigma bumps are not worth talking about while one-half of three-sigma claims in the literature are wrong.

But even if this new source of CP-violation is real, it is just completely insane to claim that it implies the existence of the fourth generation. Most theories of "new physics" beyond the Standard Model - and supersymmetry in particular - predict new sources of CP-violation. The fourth generation is just a very marginal possible answer.

Sociological summary

I think it is very bad that the "real science" is being increasingly influenced by predetermined stories by journalists who start by deciding what the story should say and then they cherry-pick, distort, or even fabricate the available evidence. It's even worse that many scientists are actually actively participating in this ethical attack against science.

And that's the memo.

1 comment:

1. You had me at "Hello".