A Discovery At the Tevatron! – Maybe

April 6, 2011 at 12:27 pm 8 comments

The CDF Collaboration released this plot today (arXiv:1104.0699, 6-April-2011):

Extra peak in MJJ distribution from CDF

MJJ distribution, after background subtraction, from CDF

The blue peak at MJJ = 145 GeV is not predicted by the standard model, of course.

The CDF paper is very clear and sober, and it is good that the collaboration reported these results. Let me outline the analysis in a few paragraphs.

The result does not come from a search for new physics (take note of this please) but rather from an attempt to measure the cross section for di-boson production in which one W decays leptonically (e or μ) and the other boson decays to a pair of jets. The invariant mass distribution includes a peak near the W and Z masses – the red peak in the plot above – as it should. The mass resolution is about 12 GeV and hence not good enough to resolve individual W and Z peaks; the W peak dominates anyway.

There is a huge continuum background from W bosons produced with two jets. At the Tevatron, these jets typically come from radiation off the incoming quarks. Here is the distribution before background subtraction:

MJJ before background subtraction

MJJ observed, and compared to SM expectations

As you can see, most events are pedestrian W+2 jets, and the excess clearly shown in the first plot above appears on a rapidly falling spectrum.

CDF have studied V + jets for many years, and have written several important papers on the jet ET and MJJ spectra. They also have a long distinguished record in QCD jet studies, so I would not suggest that they simply don’t know how to model the W+jets background. But one can see, looking at the second plot, how challenging this analysis is.

The elements of the analysis are simple: select a good quality, central, isolated electron or muon, and veto an event if there is a second lepton. Ask for a missing transverse energy MET > 25 GeV, to account for the neutrino from the W decay. Reconstruct jets with the CDF standard methods (cone algorithm with ΔR = 0.4) and require ET > 30 GeV and |η| < 2.4. Furthermore, the di-jet system must have pT 30 GeV and |Δη| < 2.5. These cuts are dictated by the properties of WV production (i.e., the original aim of the analysis), and do not sculpt the MJJ spectrum above 100 GeV, according to the article. Nothing weird, tricky or obscure, here.

As already stated, the main background is W+jets, which CDF simulates using ALPGEN+PYTHIA. The normalization of this component comes from fitting the MET distribution. The minor backgrounds are harmless and are normalized by theoretical and measured cross sections.

The initial fit, with no extra peak (red curve in the first plot above), does not describe the data in the 120-160 GeV mass range. For that regions, a Kolomogorov-Smirnov test gives 6×10-5, which is a very small number (roughly speaking, a p-value). So speculation as such is motivated.

The CDF physicists did the obvious and reasonable thing – they included an extra Gaussian in the fit to the MJJ spectrum, fixing its width according to the known di-jet mass resolution (13.5 GeV) and allowing the peak position and the height to float. Not surprisingly, they get a good fit, as you can see in the first plot.

They did a careful job evaluating the statistical significance of the peak. No smoke and mirrors here, happily. They took Δχ2 as their statistic and used a toy MC technique to evaluate the significance of the peak, including the look-elsewhere effect. They included systematic uncertainties and found that the probability to observe an excess larger than what they see in the data is 7.6×10-4, corresponding to 3.2σ. If they leave out the systematics (the most important ones are the W+jets renormalization scale, the jet energy scale and the shape of the QCD multijet background), they obtain a probability that is more than seven times lower.

Of course a number of cross checks were done. The excesses in the e and μ channels are compatible. Together they amount to about half of the WV signal, which itself validates the analysis. The physicists tried altering the background shapes by changing the theoretical parameters in their simulations. They varied kinematic thresholds. They tried reweighting the simulation according to the observed RJJ distribution. This quantity is a measure of the angular separation between the jets, and hence is highly correlated with MJJ. This test is a bit ambiguous so the question of a mis-modeling of the di-jet angular correlations might not be completely closed. But in any case the excess does not disappear.

Other obvious things to look at: There is no excess in Z+jets. And the jets here in the W+jets sample don’t appear to be b-jets. They simulated a generic di-jet resonance with a cross section of 4 pb, and found that changes in the spectrum as jet thresholds were varied are consistent with the data. There is no structure in M.

So we have a nice, statistically-significant bump in this mass spectrum. Some theorists already posted their hypothesis (Buckley, Hooper, Kopp and Niel, arXiv:1103.6035, 31-Mar-2011) — even before the CDF Collaboration released their results! (I agree completely with Gordon Watts’ comments about this.) You can also read a report in the New York Times. I’ll not comment on that.

Of course, the observation requires confirmation. I am very sure that physicists at both ATLAS and CMS are studying their data carefully. Most probably, the D0 Collaboration will try to say something soon about it, too. So let’s pay attention and keep an open mind. Speculation is fun, but fluctuations do come and go….

(Disclosure: I am in inactive member of the CDF Collaboration but played no part in this analysis. I devote all of my research time to CMS.)

Other bloggers have already commented, ahead of me: See Physics and Physicists, Not Even Wrong, The Reference Frame, A Quantum Diaries Survivor and Cosmic Variance. I’m sure there will be lots of discussions on these blogs – and if I am lucky, a little bit here, too. ūüėČ


Entry filed under: Particle Physics.

Impressive New Results from LHCb Top background shapes and the CDF MJJ Anomaly

8 Comments Add your own

  • 1. Spencer Chang  |  April 6, 2011 at 1:24 pm

    Michael, why isn’t it a little bit worrisome that the original diboson analysis saw no significant bump at 150 GeV, as shown on the public web page (in the electrons and muons combined result plot)?

    The progression in the published thesis is to notice a 150 GeV discrepancy in that plot (which I don’t particularly see). So by that wording, today’s analysis is not a blind analysis, which has an enhanced 150 GeV bump.

    Some of the steps in the analysis seem a bit arbitrary to me (like changing to absolute JES, going to exclusively 2 jets, raising the jet E_T requirement). The claim is that the higher jet E_T cut won’t sculpt a mass bump at 150 GeV, but it certainly will sculpt the low mass data which is important for the background fit.

    In particular, if the predicted W+jets background didn’t have that funny shape at low m_jj, that you could have more of the other backgrounds (which have longer tails) come into the fit and reduce the high bump.

  • 2. Michael Schmitt  |  April 6, 2011 at 1:48 pm

    Hi Spencer,

    you make a good point – this is not a blind analysis, and there is a kind of bias because someone noticed something – this is surely subjective. But that’s also how the Upsilon was discovered, I believe, so we have to live with this bias. The best way forward is to see what other experiment get when they do a similar analysis. This also addresses, to some degree, the question of sculpting.

    Although I am not an expert in this analysis, I believe raising the ET threshold to 30 GeV is justified because one cares about the shape at 150 GeV, and one wants to suppress QCD backgrounds. I don’t see that this simple variation produces a Gaussian peak directly, though any arbitrary change in the analysis may enhance or diminish the peak/fluctuation. So there is danger that the person doing the analysis keeps only those changes which enhances the peak, thereby going from essentially nothing to 3-sigma. I really don’t think this is happening here, though.

    The poor agreement below the WV peak is not nice, for sure, but my guess is that this reflects trigger thresholds, and has little impact on the background normalizations.

    thanks – it is good to get your comments !!!


  • 3. Suspicious Bump « Not Even Wrong  |  April 6, 2011 at 5:21 pm

    […] For other blog postings about this well worth reading, try Michael Schmitt, Resonaances, Gordon Watts and Flip […]

  • 4. Matti Pitk√§nen  |  April 6, 2011 at 11:54 pm

    TGD suggests two explanations for the possibly discover new boson. The first explanation is in terms of an octet of exotic weak gauge bosons predicted by topological explanation of family replication phenomenon. This explanation fails because the boson prefers decay to quark pairs.

    Second explanation is in terms of 15 year old of a scaled up copy of hadron physics whose proton would have mass around 512 GeV. I talked about this with Tony Smith with long time ago! The pions of this physics would be produced abundantly and charged pion would decay to W and quark pairs in accordance with the basic observation. The mass of the pion would be by naive scaling 71.7 GeV but p-adic scaling by factor two is possible and produces 143.4 GeV mass: Lubos mentions 145 GeV mass as the most probable estimate.

    For details see my blog posting

    Matti Pitkanen

  • […] abr. 2011): M√°s entradas sobre este tema aqu√≠,¬†aqu√≠, aqu√≠, aqu√≠, aqu√≠, aqu√≠, aqu√≠,¬†aqu√≠, aqu√≠, aqu√≠¬†y aqu√≠. Obviamente, al final s√≥lo ser√° una simple fluctuaci√≥n estad√≠stica… de la […]

  • 6. Suspicious Bump | Yonkers Mesothelioma Lawyer  |  April 10, 2011 at 8:32 am

    […] For other blog postings about this well worth reading, try Michael Schmitt, Resonaances, Gordon Watts and Flip […]

  • […] I mentioned during a particle physics lecture today that sometimes big results grow from small statistical indications, but more often than not these turn out to be false detections. I wonder what this will turn out to be? The CDF Collaboration released this plot today (arXiv:1104.0699, 6-April-2011): The blue peak at MJJ = 145 GeV is not predicted by the standard model, of course. The CDF paper is very clear and sober, and it is good that the collaboration reported these results. Let me outline the analysis in a few paragraphs. Th … Read More […]

  • 8. stringph  |  April 13, 2011 at 5:57 am

    There is now a very simple explanation in terms of a slightly (few percent) misestimated dijet energy :


    This shows us that systematic errors can be MUCH more significant than statistical!


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Trackback this post  |  Subscribe to the comments via RSS Feed

April 2011

%d bloggers like this: