CDF W JJ Resonance Closer To 5 Sigma Now
    By Tommaso Dorigo | May 31st 2011 05:21 AM | 33 comments | Print | E-mail | Track Comments
    About Tommaso

    I am an experimental particle physicist working with the CMS experiment at CERN. In my spare time I play chess, abuse the piano, and aim my dobson...

    View Tommaso's Profile
    The fact that I am swamped by the too many activities I am involved in these days can be gauged by things like the following: I get to know about important new physics results coming from an experiment I am part of by... private communications from amateurs! Knowledgeable and informed ones, of course -but that's not the point.

    So the news I got from Tony Smith today is that the jet-jet resonance unearthed by CDF in W+jets events, the one which made everybody cry Higgs just one more time about two months ago, is still there once data a significant amount of new data (three more inverse femtobarns, i.e. an additional 70%) is added to the analysis. The result which was shown by Giovanni Punzi at the "Rencontres de Blois" last week is now based on a total of 7.3 inverse femtobarns of proton-antiproton collisions collected by CDF in Run II. The added data confirm the presence of a suspicious, quite suggestive hump in the otherwise falling mass distribution of pair of jets produced in association with a W boson, as shown below.

    The left panel shows the mass distribution, with data (black points) compared to contributing processes (see legend for details). The right panels shows the excess data with respect to all background processes -i.e. W+jets, top, Z+jets, and QCD. The WW/WZ signal is prominent, but the additional excess at about 150 GeV of mass is unexplained, unless one marries the hypothesis that a new huge resonance (in blue) is contributing.

    If you are new to the analysis which produced the figure above, I will summarize it in as few words as I can. CDF was searching for semileptonic WW/WZ production, by identifying a leptonic W boson -i.e. one decaying to an electron-neutrino or muon-neutrino pair- and then picking the two most energetic jets produced in association with that particle, tuning the selection cuts in order to evidence the small signal of a W->jj or Z->jj resonance. This game can be played despite the small signal-to-noise ratio of WW/WZ production with respect to the W+jets QCD background, provided that the dataset is large and that one has a good description of the kinematics of background processes.

    The authors found the WW/WZ signal, but noticed that the background shape at higher masses was poorly described by their model. They realized that a Gaussian signal lying on top of all backgrounds could explain the feature away, and proceeded to study the effect. They could not explain away the observed feature by known systematic sources (such as jet energy scale -the obvious suspect- or other details of background modeling). So they set out to publish the effect as a hint of a new particle signal.

    That the effect was not a statistical fluke but a systematic effect was quite evident from the beginning, so the fact that the added data now confirms the earlier observation, adding units of discrepancy to the original significance, is absolutely not a surprise. What I am surprised about is that CDF produces the updated spectrum but does not provide more comments than those present in Punzi's slide above. In other words, CDF is still investigating, but in the meantime they "confirm" the previous effect. What should the end user conclude ? Is this the Higgs ?

    No, it is not the Higgs. And it is not a new particle. It is, in my humble opinion, a problem in the modeling of backgrounds, one which was unnoticed before only because it is small enough to have escaped previous attempts at "tuning" the simulations. Punzi spends a couple of slides of his interesting talk at Blois to discuss the possibility that the hump is a mismodeled top quark background, but concludes that it cannot be top. I however continue to think that the problem lies in the W+jets background...

    Time will tell, but not CDF time -if CDF just updates the result by throwing in more data, little is learnt. Instead, the other players will come in and show whether they also see a hump at 145 GeV. DZERO, ATLAS, and CMS all have already collected datasets large enough to prove or disprove the CDF effect. Note, however, that if the mismodeling is coming from a Monte Carlo simulation, and if all experiments use the same, they are likely to also see an effect! That is why I hope that more time will be spent on understanding how reliable are the simulations in reproducing the kinematics of jet radiation accompanying vector boson production. Let us look at Z+jets, gamma plus jets, 1,2,3 and more jets. Let us check the delta R distributions; let us verify how good an agreement we have in those spectra. I have seen little of this -but it might be that, as I noted at the start of this post, I am distracted by other things!


    How can 0.5 inverse fb of data collected by Atlas and CMS so far possibly suffice to prove or disprove the effect seen by CDF?

    Dear Anon,

    1) CMS and ATLAS are better detectors. They have much extended lepton coverage with respect to CDF (parts of which, used in the Wjj analysis, are 25 years old). Extended lepton coverage means multiplicative factors in signal acceptance; better calorimetry means higher resolution in Mjj, which results in higher signal-to-noise ratios. To give you a semi-quantitative example on the acceptance issue in multi-body final states, imagine you trigger on leptons in the rapidity range [-1:1] in CDF and [-2.5:2.5] in a LHC experiment, and that you trust your jets in the range [-2:2] in CDF and in [-3:3] in a LHC experiment. The former is a x2 factor in acceptance (grosso modo), the latter is a x1.5 factor that applies to each jet. You thus get an enhancement in signal by a factor 2x1.5x1.5=4.5.
    2) The cross section for diboson production (the reference process for this search) is an order of magnitude higher in 7 TeV proton-proton collisions than it is in 2 TeV proton-antiproton collisions. This means not only that 0.5/fb can equate 7/fb in signal content, but also that signal to noise may be higher in the LHC (the latter is actually not a straightforward point, but I mention it for completeness).

    1) and 2) alone mean that 0.5/fb of LHC data surpass the sensitivity of 10/fb of Tevatron data on these diboson processes. But there are other factors to consider. One is manpower: CDF has now of the order of 300 active members, the LHC experiments have 3000 each. Another is that different experiments suffer different systematics: the one CDF suffers here might just not be present in the LHC analyses.

    Many thanks for your enlightening relply. So we are looking forward to the conferences :-)

    In the last 2 months the background estimates have been looked at carefully (see e.g. It has been estimated at LO, NLO, LO+matched shower (several monte carlo's) and uncertainties have been estimated. It is becoming more and more unlikely this is a background mismodelling. What is left is some subtle detector effect.....
    That said CDF can look at the background estimates by lowering the jet cut to eg 20 GeV (QCD W+2jet production is dominating while the bump significance drops below 2 sigma) and raise the jet cut to eg 40 GeV (top production W+2 jet is dominating) to demonstrate they understand the background modelling.

    at Tevatron (ppbar) 1.98 TeV the ratio between the cross-sections of W+jets/VV is in the order of 5, while at the LHC (pp) 7 TeV the ratio is ~20. With 36/pb ATLAS could see nothing due to the smallness of the signal compared to the sistematic uncertainties.

    But perhaps, with 500/pb...

    Thanks R2, I will take your numbers as good. You could then also tell us what the absolute numbers are - I suspect the cross section for the signal is higher by over a factor 10 (for ttbar it is 18, for single bosons it is only about 5 or so if I'm not mistaken). Given the acceptance improvements, and the better Mjj resolution, I insist that the CMS and ATLAS detectors just need 500/pb to place a limit on Wjj which excludes the CDF signal.

    Strictly speaking in 2) you're comparing the backgrounds. If you compare the signal you must specify the model able to produce it but presumably LHC performs better at that too.

    You suggest we now look at look at Z plus jets, gamma plus jets, W plus jets simulations... Well this is what quite a bit of theorist have been doing for a living well before all this came up. They've looked harder still since the Wjj excess showed up (see I'm afraid this cannot be thrown purely into the theorists court. If this is not new physics it is either a subtle detector effect (perhaps combined with some theoretical uncertainties). In either case, a systematic effect or new physics, this is very important to understand.

    No, not in the theorists' court. I suggest that detailed studies of the kinematics of similarly behaved processes be performed, by comparing data with simulations performed similarly to the ones used to get the background shape in the Wjj analysis. I think we agree.

    Hi Tommaso, wasn't it true that the excess in Wjj only occurred in events where W decayed to the neutrino pair, and not the electron pair? Or do I misremember? What happened with this strange asymmetry that may invalidate the claim that it's just an "ordinary W" existing over there as an intermediate state? Is it now seen with the electron pair as well?
    Careful, Lubos, these are W's, not Z's, so they produce a charged-current process whereby you get either an electron-electron neutrino pair, or a muon-muon neutrino pair, or a tauon... you got me.
    However no, the signal they got was from combining the ev and mv final states. It was nicely consistent in the two datasets, and as far as I know it still is.

    You seem to be right. I must have confused it with one of the other, less real bumps.
    In that case, I don't understand what you mean by "mismodeling the W plus jets background". these are just words. is it monte carlo that you do not trust? the CDF detector simulation? (your detector, I believe). GEANT?

    In that case, I don't understand what you mean by "mismodeling the W plus jets background". these are just words. is it monte carlo that you do not trust? the CDF detector simulation? (your detector, I believe). GEANT?

    Perhaps this would get sorted out faster if CDF devoted more resources to studying the excess and looking in correlated channels instead of working to show that the precious Higgs limits are safe.

    Last time I checked you are still a CDF collaborator so please enlighten us with all your good ideas (DeltaR, oh boy nobody had thought about that! )
    Also might be good to just browse through some internal pages, public ones and I'm pretty sure you can find all the checks you quote and many more.
    People here are working hard and certainly do not deserve to be told on a blog, from somebody that does not take the time to even read notes from his own collaboration, that they are not.

    Hey, If you are a colleague why don't you leave your name here? Anyway, I have no good ideas. Please do not lose sight of the fact that this is an outreach site, not a place from which I pretend to be giving advice to analysts. I would ask for more money for that! Sorry for not getting in details, I am in Athens with no time and no wifi right now. Later, T.
    147(5)GeV - Excellent! This shifts the center of the data to my prediction (148 GeV) in eq. 18 of

    Daniel de França MTd2
    Could that be a 4th generation of a down decaying?
    As to eq. 18 of J Gregory Moxness, it says
    "... the mass of the Higgs boson ... = 147.98904797 GeV/c2 ..."
    as Tommaso said in his earlier (6 April 2011) post about the CDF feature at 120-160 GeV with "center of the data ... 147(5) GeV":
    "... it cannot be a Higgs boson,
    because if a Higgs boson were sitting at 140 GeV with such a large production rate we would have seen it decay into WW pairs a long time ago ...".

    Also, over on Resonaances, Chris said:
    "... this looks most like a slight offset in energy scale ...".

    In slide 32 of his Blois presentation, Giovanni Punzi said:
    "... What happens if we change the Jet Energy scale?
    Result of the fit scaling JES up by 7 per cent ...
    always above 3 sigma ...".

    Chris also said:
    "... around 210 GeV this feature repeats ...".

    My guess is that the feature around 210 GeV is too low in statistical significance to be presented as evidence or even as very interesting.

    My selfish interest, based on my 3-state Tquark - Higgs model described on my web site (click on Truth Quark link on the index.html page), would be for (in addition to the well-known 174 GeV Tquark middle peak):

    a 120-160 GeV peak corresponding to a low mass Tquark state
    a 210 GeV peak corresonding to a high mass Tquark state.

    a problem for my model is stated by Giovanni Punzi in his slide 33 where he says:
    "... There is no significant tagged component ...".

    As to how much tagged component there is, according to a CDF note update at
    "... one ... jet ... Loose tag ...[in]... Excess region ...
    rate in ... muon ... samples ...[ is about 15 per cent ]
    one ... jet ... Loose tag ...[in]... Excess region ...
    rate in ... electron ... samples ...[ is about 13 per cent ]...".

    Maybe all will become clear when D0 and CMS and ATLAS get their results.


    Yes, while my paper states the particle as being Higgs, the model I use to determine the mass is clearly not specific to the Higgs mechanism. Please focus on the model used to make the prediction. My prediction may be a Higgs sector-like particle (i.e. Lisi Higgs sector particles) or even a mechanism outside SM.

    Bottom line - it is a specific prediction (not retrodiction) and the model used does not (yet) conflict with the experimental evidence.

    I think that peak is partially caused by jet merging. That can produce shadow peaks like that when the jets in a process have a long p_T tail. In that case the high mass peak should consist of a high mass merged jet and a small that was just picked up. Many experiments use the one-jet-size-fits-all philosophy, a jet size that was tuned to a certain energy scale. I admit, they use more fat jets for the QCD sometimes, but it is still a two-size-fits-all. I was not satisfied with the study of the systematics in that article, though it still can be true. I'm just not convinced, and if the authors can come up with conspiracies of the source of the peak, I'm allowed to come up with conspiracies too for things that went wrong :)

    What conspiracies are you referring to? The CDF authors simply selected W+2jet events, plotted the di-jet invariant mass and observed an excess bump relative to the background modelling.

    Sorry, I was not clear. I used the term conspiracy for the explanation, that this bump can be explained by a leptophobic Z' that has so many new coupling constants.
    Whereas the jet merging is something that is usually not taken into account. The first peak on the plot is associated with WZ and WW productions, where W->lnu and W/Z->jj, and the jet pair gives the invariant mass peak. Unlike the QCD dijet, these W/Z jets are coming from a relatively high pt objects and, in extreme case may produce merged jets. Thats obviously just one heavy jet, but there may be additional jet picked up from noise,underlying event or pile-up. Combined with the heavy jet, it forms an higher invariant mass than the heavy jet alone, so it is like a shadow of the original peak.
    The amplitude of the second peak surprises me though, but as it is in the same order of magnitude as the WZ/WW or tT production, I'd bet it is a systematic effect.

    Ok, in that case I wholeheartedly agree with the conspirancy comment. Given we have 3 papers looking at the SM as a possible explanation and order 50 papers giving BSM explanations says everything.... If it was the reverse we would by now have understood the excess in terms of the SM or be more confident this could be BSM physics.

    Thanks for explaining the jet merging a bit more, I certainly did not consider this possibility. You seem to have thought about this, so a few quick questions:
    1. Why is this not simulated by alfgen+mlm matched pythia or by sherpa?
    2. The azimuthal cut dependence as shown on the CDF web page suggests the events in the bump consists mainly of back-to-back jets. If the second jet is picked up from noise or underlying event why would it have this correlation?
    3. CDF shows also the dependence of the dijet mass on the jet balancing cut (Et1/Et2). As your second jet is relatively soft, why would the bump seem stable under this extra cut?
    4. The excess events in the bump have the kinematic feature that the Pt(W) (or Pt(jj)) is cut-off at 100 GeV. How is that consistent with the higly boosted W assumption (This edge in the Pt distribution is very suggestive of a 300 GeV particle decaying in a 80 GeV particle and a 150 GeV particle.)

    Not sure I understand the jet merging reasoning. It seems to me this would then also affect the di-jet invariant mass distribution (i.e. select 2 jet events without a W). This dijet invariant mass observable agrees very well with the background modelling.....

    I think here is not conclusive comparing with a plain dijet invariant mass; missing ET requirements could be quite different (I would say in the opposite) so, for sure, you are looking at different processes.


    Agreed... It would not be conclusive. However, there is an large body of data out there which has been compared to the theoretical calculations. With this data as a guide, enormous progress has been made in the last 20 years on how to make quantitative prediction for collider observables. We are beyond making qualitative QCD predictions using some parametrized model or re-weigthing the theory using the data.
    Assuming you can just change something in eg Sherpa to get rid of the CDF excess is not going to work. Such a change will affect all prediction, at CDF, D0, ATLAS and CMS. This makes understanding the CDF excess so important. We might be missing something subtle in the theory predictions affecting all experiments.

    my thought was more about jet merging. I would say that jet merging is different with large or small MET .
    but maybe I am wrong. (in general the number of "low" pt jets is correlated to MET)



    I'm one of those amateurs - at the edge of physics.

    Lee Smolin has us just outside the 'group think' camp.

    Perhaps the 'Trouble with Physics' is that amateurs are the canaries in the coal mine.

    Professionals are drowning in data on this one.

    Help us out here - tell us what's really going on between the outhouse and a shed - in the fb−1?

    Is there any possibility of this being a light pseudoscalar Higgs boson?

    Remembering this paper form the past:


    Hi Paul,

    not clear what I can do for you. The paper is not from the past, it is from the pleistocene...

    As for new SUSY particles emerging from 1/fb datasets, forget it. But forget it also from 10/fb.