• John
  • Felde
  • University of Maryland
  • USA

Latest Posts

  • James
  • Doherty
  • Open University
  • United Kingdom

Latest Posts

  • CERN
  • Geneva
  • Switzerland

Latest Posts

  • Aidan
  • Randle-Conde
  • Université Libre de Bruxelles
  • Belgium

Latest Posts

  • Vancouver, BC
  • Canada

Latest Posts

  • Laura
  • Gladstone
  • MIT
  • USA

Latest Posts

  • Steven
  • Goldfarb
  • University of Michigan

Latest Posts

  • Fermilab
  • Batavia, IL
  • USA

Latest Posts

  • Seth
  • Zenz
  • Imperial College London
  • UK

Latest Posts

  • Nhan
  • Tran
  • Fermilab
  • USA

Latest Posts

  • Alex
  • Millar
  • University of Melbourne
  • Australia

Latest Posts

  • Ken
  • Bloom
  • USA

Latest Posts

Flip Tanedo | USLHC | USA

View Blog | Read Bio

A hint of something new in “W+dijets” at CDF

Update: for those who read this in time, there will be a seminar on this result broadcast online through the Fermilab webpage. (The talk will be at the level of physicists working in the field.) This result was also mentioned in the NY Times today.

Even though its running days are numbered, the Tevatron reminds us that it can still muster up an interesting hint of new physics. This is a quick post on a brand new result from the CDF collaboration, “Invariant Mass Distribution of Jet Pairs Produced in Association with a W boson in ppbar Collisions at Sqrt(s) = 1.96 TeV” [arXiv:1104.0699]. Just to whet your appetite, here’s one of the plots that we’d like to understand (hot off the press!):

Unpacking the title

Before diving in, let’s first understand what the title means. It’s somewhat cumbersome, but that’s because it encodes a lot of good physics. It turns out that it’s easier to start at the end of the title and work our way backwards.

  • “…in p-pbar Collisions at Sqrt(s) = 1.96 TeV
    This is telling us about the general Tevatron experiment: they collide a proton (p) with an anti-proton (p-bar) at an energy of about 2 TeV. A TeV is roughly the kinetic energy of a flying mosquito. In Feynman diagram calculations it’s often useful to use the parameter s, which is the square of the energy, rather than the energy itself, hence the weird “Sqrt(s) = 1.96 TeV.” For comparison, recall that the LHC is now operating at 7 TeV and will eventually go up to 14 TeV.
  • “…in association with a W boson …”
    This should sound a lot like the “associated production” diagram that we drew for the Higgs boson in a previous post:Reading from left to right, we have a quark and an antiquark producing a gauge boson (in this case the W boson) along with a Higgs. I’ll tell you right now: the CDF result does not appear to be the Higgs, but if there is a new particle responsible for it, it is produced from the same diagram (with the dashed line representing the new particle). [Correction, 6 Apr: a few people have correctly pointed out that the new state could also come from a “t-channel” diagram with an intermediate fermion. (I’ll leave the actual diagram as an exercise for those who have been reading my Feynman diagram posts).] By the way, the W decays before it reaches the detector. Two signals are W?e? or W???, which appear in the detector as a charged lepton and “missing energy” (the neutrino).
  • “… Jet Pairs Produced
    We also know that the Higgs decays into other stuff before it reaches our detectors. Thus we have to tell our detectors to keep an eye out not for the exotic particle, but for the ordinary stuff that it decays into. If it decays into quarks or gluons, then we know that we end up with jets. Thus this paper is looking for a W boson (which itself becomes a lepton + neutrino/missing energy) and a pair of jets. This is the criterion for picking out ‘interesting’ events relevant for this analysis.
  • Invariant mass distribution of…
    Now that we’ve picked out interesting events, we want to plot them in a way that can tell us if there’s a new particle hidden in the data. The easiest way to do this is to go “bump hunting,” which is what we discussed for the detection of the Z boson. The key idea is this: if we sum the energy and momentum of the two jets, we should get the mass of the intermediate particle that produced them (if they were produced by the same particle). This sum is called the invariant mass, and by plotting the number of interesting events based on the invariant mass, we can look for bumps that are characteristic of new particles.

Phew—that was quite a lot packed into a title. But now we’ve established most of the physics to understand the plot and see what people are [cautiously] excited about!


Here’s the plot, once again:

On the horizontal axis is the invariant mass of the two jets, which is roughly the sum of the jet energies. The vertical axis is the number of events in the data set with the given invariant mass. If there were a particle which produced the two jets, then there should be a bump in the number of events with an invariant mass around the mass of the new particle.

So… what’s all of that colorful mess in the plot? It’s the thing that makes experiments hard: background. We have identified a particular experimental signal (2 jets, a lepton, and missing energy) which could come from a new particle. What we still have to account for are “boring” processes which could lead to the same signal. By “boring” I mean Standard Model processes that we already understand. Here’s the inventory from the plot above:

  • The red contribution to the histogram are events where a pair of Ws or a WZ pair are produced. The second W (or the Z) then decays into a pair of jets.
  • The big green contribution comes from various processes where a single W is produced and the two jets are separately produced independently of the W.
  • The white (with a pink border) sliver are events where a top-anti-top pair are produced. These tops each immediately decay into b (or anti-b) and a W boson. One of these W bosons decays to lepton + neutrino and is tagged by the experiment, while the other one might decay into jets. Now we have four jets (2 from the b quarks, two from the decay of a W), but it is possible for two of those jets to get ‘lost’ because they don’t fulfill the detector criteria for identifying jets. (This is a notoriously subtle thing.) There is also a contribution from the production of a single top.
  • The blue sliver is the production of a single Z boson with two jets. The Z decays into two leptons One of the leptons can be ‘hidden’ because the particular search only looks at particles which are fairly perpendicular to the beam direction, or misidentified as a jet.
  • Finally, the shaded sliver is QCD background: these are gluon mediated processes that can give two jets and a lepton.

Once we take all of these background effects into account—and this is a very nontrivial thing to do—we can subtract these from the actual number of observed events per invariant mass bin. This “background subtracted data” is plotted below:

Now things look rather interesting. First note that not all of the background is subtracted: they leave in the WW and WZ background because these will also produce a characteristic bump (red line) because the 2 jets come from a single particle (indeed the bump is around the W and Z mass). The other backgrounds have broad, smooth profiles and can be reliably subtracted—bumps are harder to subtract so we keep them in. (Update: I’m told that this may also be partly included for comparison reasons: the W/Z bump is really well understood, so it helps to be able to use it as a measuring stick.)

What’s particularly neat, however, is that there seems to be a second bump with a peak right around 150 GeV. This is what is shown in the blue line. The significance of this bump is around 3.2 standard deviations, which roughly means that we can be 99.7% sure that this is not a statistical fluctuation.

It’s not the (standard) Higgs

If this bump really does come from a particle with mass around 150 GeV, then the first thing one might think is that this is the first hint of the Higgs boson. Indeed, we even showed above that the production of the Higgs includes diagrams that would give this particular signal when the Higgs decays into two quarks. However, one very, very interesting part of the analysis is that it does not seem like this bump could come from the standard Higgs boson!

The reason is simple: we understand the standard Higgs well enough to know that if it had a mass of 150 GeV, then we would expect an effect (a bump) that would be about three hundred times smaller. In the parlance of the field, the observed bump corresponds to a particle with a 4 picobarn dijet cross section, while a 150 GeV Higgs is expected to have a 12 femtobarn dijet cross section.

Further, CDF has already done a closely related analysis: WH ?l? b b-bar. This is basically the same analysis as the present paper, except that they are able to identify jets that come from b-quarks (this is called b-tagging). The analysis with b-tagged jets showed that there was no significant excess in the range 100 — 150 GeV.

What this means is that if this bump is indeed coming from a new particle, then it must not be a particle which decays into b-quarks, at least not very often. We know, however, that the standard Higgs does decay into b-quarks, so this hypothetical new particle could not be the usual Higgs.

This is actually much more interesting, since this could either suggest a non-standard Higgs sector or it could be a sign of completely different new physics.

I should note that there is one sentence whose significance is a little unclear to me:

We compare the fraction of events with at least one b-jet in the excess region (120-160 GeV) to that in the sideband regions (100-120 and 160-180 GeV) and find them to be compatible with each other.

Basically, they look at how many b-quark jets were in the bump versus those that weren’t in the bump, and they find that the number is roughly the same. This seems to imply that whatever is causing the bump is not decaying into b quarks, but I’m not an expert on this and might be misreading it.

Where to go from here

Don’t get too excited, though. Nobody is breaking out champagne bottles yet. Three standard deviation effects have been known to come and go—i.e. it is possible that it is just an unlucky statistical/systematic fluctuation. For example, it might be a mis-modeling of the background that had to be subtracted. (The three standard deviation significance assumes that one “knows how to estimate what one doesn’t know,” as one person explained it to me.) All the same, I expect that there will be plenty of model-building papers by eager theorists in the next few weeks. [By the way, CDF has known about this effect for some time; the current excitement comes from breaking the 3 standard deviation significance and their subsequent publication of the result.]

There are a few things to look out for:

  1. More data! We measure data in “inverse femtobarns” (1/fb). The current paper is based on the analysis of 4.3/fb. My [outsider’s] understanding is that CDF should have around 10/fb by the end of the year, so the collaboration should be able to say something with more significance if this is a real effect.
  2. What about D0? Fermilab’s other collaboration should be able to corroborate (or refute) this effect.
  3. I do not believe that the LHC has enough data to say much about this at the moment, though I understand that we could be looking at 1/fb of data by summer time, and maybe a few inverse femtobarns through 2012. If the signal is real, there might be some hope to see the effect before the long shutdown at the end of 2012.

There are a lot of people who are cautiously optimistic about this. It’s almost certain that many theorists will jump on this to see if their favorite models can be tweaked to give a 150 GeV particle decaying to jets (but visible in the b-jet analysis), and that’s part of the fun. I look forward to seeing how things develop (and perhaps jumping in if the opportunity presents itself)!

Acknowledgements: I would like to thank my experimental colleagues, SP and DP for many helpful conversations. Any errors in this post are purely due to my own misunderstanding. I challenged some hep-ex grad students to foosball to try to squeeze info out of them before the paper was published… but I lost and they didn’t spill any beans. [They’re also strictly prohibited from such gossip… especially to theorists.]

  • Matt Reece

    Hi Flip,

    This is a nice post, but I have a couple of comments. You wrote:

    we understand the standard Higgs well enough to know that if it had a mass of 150 GeV, then we would expect about three times more events than were found.

    Can you explain this estimate? Here’s my attempt, which gets a rather different answer: eyeballing Fig 3.3 in the big Djouadi review hep-ph/0503172, it looks like the HW cross section at the Tevatron is ~ 50 fb. After asking for the W to go to e or mu, this drops to 10 fb. The data sample was 4 fb^-1, so this is about 40 events, and acceptance effects will cut that down by another factor of 3 or so. But the plots show about 200 events in this bump. So I would say that for a standard Higgs we expect about a factor of 20 fewer events than were found. (Not to mention that a 150 GeV SM Higgs wouldn’t dominantly decay to dijets; asking instead about an MSSM Higgs that would, though, doesn’t help improve the rate.) This is back-of-the-envelope, but I would have to be missing something important for it to be off by much more than a factor of 2 either way.

    the CDF result does not appear to be the Higgs, but if there is a new particle responsible for it, it is produced from the same diagram (with the dashed line representing the new particle).

    Or, it could be a t-channel process with a quark radiating both a W and the new particle. A few people have already posted such things on the arxiv. (Unfortunately, UA2 didn’t have enough luminosity to exclude a light Z’ coupling only to hadrons, and our current colliders are too high in energy, so that a simple dijet bump search is swamped by QCD background at this low mass.)

    Best wishes,

  • Joao

    Hi Flip

    Great article. You’ve really nailed it to the point and made it really understandable for “normal” people like me. Looking forward to reading more from you.

  • Thanks for a really accessible explanation of all the technospeak, even I understood it. Also a really exciting result to follow up as more data comes in. I reckon it could be a really exciting year for particle physics and cant wait to see how this develops.

  • Hi Matt! Re: your first point, this was a misread on my part that I fixed about an hour after posting. (You were reading this blog between midnight and 1 am? ^_^) I misread the paper, which said that the bump was 4pb, whereas a 150 GeV SM Higgs would be have a dijet cross section of around 12fb.

    Thanks for the correction with the t-channel. I’ve made the correction in the post.


  • Nikola Nikolov (ORNL)


    Is there 5 sigma condition for a new particle discovery?



  • Jonathan Clift

    I’m just a lay person, with no real knowledge of this sort of thing (just in case there should be any smidgen of doubt in your mind about that), but if you handed me the plot with the background subtracted and explained that the resonance at around 80GeV/c^2 related to the WW and WZ cases, I would have assumed that the higher peaks were ‘harmonics’ of that, ie the situation where 160 would sometimes give two cases of 80, rather than being a totally new particle. The fact that there also seems to be something going on above 200 would lend weight to that. Also, I would have speculated that the combined energies of the WW and WZ were slightly different because of the way the ‘harmonics’ were diverging (which seems to be true if I look up the masses of the W and Z bosons on Wikipedia).

    Obviously I’m wrong, but what is it I’m missing? Why does it have to be a new particle?

  • Hi Nikola, yes I believe 5 sigma is considered a ‘discovery’. Caveat: the 5 sigma means that the bump is real… we also have to be sure that it’s not due to something more mundane. For example, the DAMA dark matter experiment has been sitting on an 8 sigma signal for a long time now, but other experiments are still trying to confirm that it’s not some systematic effect.

    Another way of saying this is that that the significance level quoted accounts for “known knowns” and “known unknowns.” If there’s something that we really don’t understand about the QCD background, then maybe such a ‘discovery’ isn’t a new particle but but some hitherto misunderstood effect in nonperturbative physics (just to throw out a random possibility).

    Hi Jon! I think I understand where your confusion is coming from and I suspect that it might come down to terminology. We use the phrase “resonance” to describe the bump in the mass spectrum, but this does not mean that we expect to see higher “harmonics” of this bump. This is a different kind of resonance than the acoustic resonance that is more commonly known. I need to think a bit more carefully about the precise relation from a physics point of view—but but in general, if we have a particle of mass X which decays into two jets, then we expect to see a bump in the “dijet invariant mass” at X, and not at 2X, 3X, etc.

    I should also say again that just because there’s a bump, it doesn’t (yet) mean that it has to be a new particle. Given the big media hoopla, I want to make it clear that the people in the field are being as conservative as possible and trying to make sure that there could be no other possible explanation.


  • Mario.

    As an engineer: if the subtracted data were real hard data, I would hold on to that and make predictions, but as these are just data sitting on HUGE pile of ENORMOUS background it could be any kind of noise and statistical randomness.

    But there is still one thing that comes to my mind.
    The obvious later onset of the first peak and even a NEGATIVE bump in the substracted data which inspires doubt over our model of the underlying “knowns”.

    The second peak is much wider and that is also what bothers me. For electromagnetic resonance – in the second harmonic you might expect double the width of the first peak, but not tripple. If it were straight data, the black wavy tail would also have been significant. Maybe a slight tweak in the underlying model is in order, apart from search for a new particle.

    Two more notes. The energy peaks could be spin-dependent, the second for double the spin of the first one — but that is only a sidenote.

    Second, I really do believe we all swim in a soup and do not realize that. My explanation of everything is that we live in a physically defined space, the kind that Dirac imagined as the first one. Gamow pictured Dirac with a dolphin as Dirac talks to him. The dolphin does not realize he is in water, that the molecules are all around him, the same way we may not realize that we live in a checkered 3-dimensional grid paper, and this physically defined grid (almost weightless) is what we may be interacting with. I explain it with the fact that even vacuum has physical properties.

    Am I wrong?

  • Hi Mario, quick responses to your comments:

    1. Indeed, the background is large and can be a very subtle thing. However, physicists work *very* hard to correctly subtract away the background (they also do more sophisticated things such as varying parameters for the background in the fit). At the end of the day we’re just counting how many events the Standard Model predicts versus how many events were actually measured. That being said, it is possible that there is something about the background that was mismodelled. Very good people are working on this aspect. We’ll have to see.

    2. *IF* the bump comes from a new particle, then the width is related to the decay properties of that particle. I’m not quite sure what you mean by ‘harmonic’ — at this level it might be best to say that this kind of “resonance” is not quite the same as an acoustic resonance.

    3. I’m not quite sure what you mean. The decay process may be spin-dependent, but the plot over energy spectrum should not be. This is conservation of energy. (Again, this is all assuming that the two jets came from a new particle.)

    4. At least within the Standard Model, the ‘soup’ that we live in is the ‘vacuum’, and as you correctly point out, our vacuum isn’t empty. In addition to quantum fluctuations, we live in a particular Higgs field background configuration. I will blog about this as soon as I get a chance (unfortunately this CDF result came up!).

    Thanks for the comments,

  • TimG

    I think this is a “resonance” in the sense that it’s a peak in the response around a certain frequency (energy being proportionate to frequency), but that the cause of this is not the same as a resonance in a classical system. In particular, if a classical system (such as a vibrating violin string) oscillates at a given frequency, it makes sense that it would oscillate at twice that frequency, as the double frequency wave would have nodes in the same places (plus extra ones). So it could satisfy the same boundary conditions. But here, it doesn’t make sense that if you had *two* particles decaying, you would see the same decay products with twice the energy. Instead, you’d expect to see *twice* the decay products (assuming you could detect them all), which I would think is a different signal.

  • Jonathan Clift

    Thank you TimG (and Flip).

    I think I get it. The problem was I was thinking of the energy on the horizontal scale as being the total energy contributed by the collision whereas you are saying it’s the energy measured for a particular ‘signal’ that has been found. Flip said the same thing in his post if only I’d read it more carefully.