Double-Parton Scattering is Not Rare

Despite lots of empirical evidence to the contrary, I tend to think of proton-proton interactions as the collision of single partons (quarks and/or gluons, one from each incoming proton) giving rise to all sorts of rich phenomena. A recent paper by Berger, Jackson and Shaughnessy reminded me that this way of thinking is too simplistic, and that the simultaneous scattering of two pairs of partons from the same two protons provides a non-negligible contribution, especially in certain corners of phase space – corners that may be quite relevant for finding physics beyond the standard model. Their paper is available on the physics archive: arXiv:0911.5348.

I grasp the essence of their paper as follows: imagine that you were looking at the production of b-quark pairs, as part of a search for the Higgs boson. You might look at the subset of events in which the hadron jets from the two b quarks are back-to-back, i.e., for which Δφ is nearly 180°. There will always be some extra activity in the event (even forgetting the contribution from other protons interacting) due to the initial-state radiation (ISR) of gluons. Naively, however, one would expect ISR to be small when Δφ is close to 180°. A standard event generator will simulate only those interactions coming from a single pair of partons, and will very rarely produce four jets – the two back-to-back b-quark jets and two other jets, which necessarily will be back-to-back, also. It would not be difficult to compare such a prediction to real data – in fact, this will surely be done as soon as the LHC delivers more data in a couple of months. According to the studies of Berger, Jackson and Shaughnessy, however, you would see a major discrepancy between the “standard” prediction and the real data…

These studies show that there would be a “surprise” contribution of jets which are themselves back-to-back, very much as if two events were overlayed. The important point is that these two “events” come from the same proton-proton interaction, and hence do not depend on the instantaneous luminosity, in contradistinction to the overlap of two proton collisions in the same beam crossing – which obviously depend on how many protons are in each bunch. Furthermore, collisions from different pairs of protons can be separated to some degree by reconstructing their different Z-coordinates (positions along the beam line), but this is not the case for double-parton scattering.

The thrust of the Berger, Jackson and Shaughnessy paper is a study showing that clear evidence for double-parton scattering can be obtained with a few pb^-1 of data at 10 TeV. Here is one of the most telling distributions of their study:

S_φ, which peaks toward one for double-parton scattering

S_φ = (1/√2) ⋅ √( (Δφ_bb)² + (Δφ_jj)²)

The quantity S_φ peaks toward one for events in which the b-quark jets are back-to-back, and the other jet pair is also back-to-back. The bulk of the distribution indeed comes from single-parton scattering (“SPS”), in which two ISR gluons accompany the b-quark jets. For the SPS component, there is no special reason why the b-quark jets should be back-to-back, or the two ISR gluons should be back-to-back; the final state populates a four-body phase space which accommodates many other configurations. For the double-parton scattering (“DPS”) component, however, we have the overlay of two two-body final states, and each individual two-body final state is necessarily back-to-back. The DPS component may be small overall, but the plot shows a very tall spike at a corner of phase space. (The SPS component is represented by the red histogram, the DPS by the blue, and the sum of the two by black histogram.)

Double-parton scattering has been investigated empirically in the past, and many papers have been written about it in the context of high-energy hadron-hadron colliders. The paper by Berger, Jackson and Shaughnessy is particularly useful and I hope that the LHC experiments will follow-up once the necessary data have been recorded. At a minimum, an overall factor σ_eff must be measured in order to make progress. On the longer term, we should try to gain some knowledge of the double parton distribution functions (see arXiv:0911.5348 for details).

The Physics in the ALICE Paper

The ALICE Collaboration published the very first physics paper on collisions at the LHC. You can see the paper on the archive (arXiv:0911.5430), and soon in the European Physics Journal.

I love to read papers reporting good, basic measurements that I don’t really understand – they give me the opportunity to go learn something new! After all, we don’t measure any arbitrary peculiar quantity because we can – every measurement has a point to it, something more than just checking a Monte Carlo event generator. So, what is the physics point behind the ALICE measurement?

The ALICE Collaboration measured the density of charged particles as a function of pseudo-rapidity (η). Here is the main plot:

Charged multiplicity as a function of pseudorapidity (η)

The horizontal axis gives the pseudorapidity, and the vertical axis gives the “density” or number of charged particles per unit of pseudorapidity.

First, let’s recall what pseudorapidity is. It is sometimes described as a measure of the angle with respect to the beam, but this is not really what is means. It’s not just a substitute for cosθ. It is an approximation to the true rapidity (y), which parametrizes relativistic boosts. In old-fashioned books on special relativity, one sees sinh(y) = βγ so one might say that y is an imaginary angle, but this certainly brings no insight. The main point of rapidity in relativistic kinematics is that a function f(y) transforms as f(y+a) under Lorentz boosts – i.e., the distribution is shifted along the y-axis by some constant amount, but the shape does not change. Hence, it is the shape which is of the essence.

Let’s write the explicit formulae:

y = ½ ln [ (E+p_z) / (E-p_z) ]

η = ½ ln [ (p₀+p_z) / (p₀-p_z) ] = -ln[ tan(θ/2) ]

So it is obvious that η is numerically close to y when the particle is relativistic and its mass does not matter, numerically. Experimentally speaking, however, η is much easier to measure, since one needs only θ, the angle of a track with respect to the beam axis. For y, one needs good momentum measurements and knowledge of the mass, which implies particle identification, which is hard.

All of this is to say that we use η because we can measure it well, but for puzzling out the physics, one should think in terms of the normal rapidity, y.

(Note that the ALICE magnet was off for these data, so they have no way of measuring momenta. For a nice synopsis of the analysis, see Zoe Matthews’ post on Quantum Diaries.)

What do you notice about the results in the plot above? The charge density hardly varies with η! Why not? Why is the density of charged tracks constant as a function of rapidity? Considering its definition as a boost parameter (see above), this should come as quite a surprise. (The small waviness reflects the small numerical differences between η and y.)

Now, to the physics. I learned what little I understand from a very nice review article by Grosse-Oetringhaus (CERN) and Reygers (Heidelberg) which came out just after the ALICE paper: (arXiv:0912.0023). The story begins with Feynman Scaling which pertains to the formation of hadrons in inclusive, inelastic processes.

Let’s begin with this picture: A stream of partons comes in along the z-axis at high energy, and new partons are produces which fly out in many directions. Do not think in terms of simple Feynman diagrams or any sort of one-particle exchange. There is just a big blob with lots of partons coming out. The key concept is that the formation of any particular final-state parton does not depend on the global picture, only the local one. The energy and flavor of the incoming partons don’t figure into the calculation – a parton is produced according to some dynamics, and that parton turns into a hadron according to a universal distribution which might depend on the particle species but not on the other partons or their kinematics. The other key idea is that the distribution functions factorize in terms of the transverse dimension (p_T) and a longitudinal quantity. The latter is expressed as a scaling variable, “Feynman x_F,” which is simply p_z/W, where W = √s is the center-of-mass energy.

Feynman posited that the probability to produce a parton of energy E goes as 1/E. He justified this on very general phenomenological grounds. From this, and integrating over p_T, and making a change of variables, one can show that the mean multiplicity N is proportional to ln(W). The total range of rapidity is also proportional to ln(W), so one infers that dN/dy = constant.

Thus, the flatness of the distribution of dN/dη is an important validation of the quark-parton picture, invented in large part by Feynman.

Here is the other important physics plot from the ALICE paper:

Increase of mean charged multiplicity as a function of √s

This compilation of measurements tests the prediction that N increases as ln(W). Clearly, this prediction fails – the increase is quadratic, not linear. This has been known for quite some time, and does not invalidate the quark-parton picture (of course). The arguments behind Feynman scaling were extended, leading to KNO scaling. KNO scaling works well but not perfectly, and the UA5 experiment showed violations of KNO scaling. It was eventually shown that the multiplicity distribution within a narrow η range follows a negative binomial distribution (NBD), or better, a sum of two of them. I can’t explain these things here (nor are they addressed in the ALICE paper), but the salient point is that these extensions of Feynman’s original line of thinking lead to the expectation that N increases quadratically with √s, as seen in the second plot above.

This piece of physics is very rich, even if not new, and it is interesting that the best Monte Carlo generators have some trouble reproducing the data accurately. For a lot more detail, see the review article by Grosse-Oetringhaus and Reygers (arXiv:0912.0023) or a good textbook on hadron-hadron interactions. For more data, just wait awhile – the LHC collaborations are only getting started!

An Excellent Start for LHC Physics (Seminar II)

Today the second public status report on the LHC was held. The presentations from the various experiments are on this public web page, and a recording of the session will be available from the CERN Document Server.

Fabiola Gianotti, spokeswoman of ATLAS, delivering the ATLAS report. Several famous physicists can be spotted among the audience.

The LHC has delivered a useful data sample for pp collisions at a center-of-mass energy of 900 GeV (roughly 20 μb^-1). There was also a very small sample at 2360 GeV (order of 1 μb^-1), which exceeds the highest energy of the Tevatron at Fermilab.

The experiments used these data to establish the performance of essentially all subdetector systems, and to demonstrate the reconstruction of physics benchmarks. Many of these benchmarks are common, so one can make a direct comparison. Here are the reported widths of reconstructed signals, in MeV:

collaboration	Ks→ππ	Λ→pπ	φ→KK	π0→γγ	η→γγ
ALICE	5.2	1.9	5.3	7.2	-
ATLAS	8.2	3.2	-	19	-
CMS	7.6	3.1	4.6	16	53
LHCb	4.3	1.4	-	11	-

.

The collaborations their capabilities to identify charged particle species. The plots from ALICE are particularly impressive, showing very clean dE/dX separation in their TPC and ITC, augmented by e/π separation in the TRD and nice separtion by TOF. They are planning a publication on the multiplicities and momentum spectra of pions, kaons and protons. ATLAS and CMS also shows dE/dX separation of e/π/K/p, and CMS used this in constructing their φ→KK signal. ATLAS also has a TDR with a very nice separation. LHCb demonstrated that their RICH detector can identify charged kaons well.

All detectors are able to reconstructed leptons (electrons and muons) well. ATLAS, CMS and LHCb showed di-muon candidates, and those from CMS and LHCb are potential J/ψ signals. It will be interesting to see these samples grow from a handful of events to clear signals, hopefully with a clear prompt and non-prompt component from B hadron decays (at higher energies).

Nice examples of di-jet and tri-jet events were shown by ATLAS and CMS, and both collaborations were able to show a jet p_T spectrum, with an excellent reproduction by the simulation. More importantly, the missing transverse energy (MET) distributions were shown and the canonical relation between MET and ΣET, again reproduced very well by the simulation. The CMS Collaboration showed results from their _Particle Flow_ algorithm, with a very good calibration and even the reconstruction of neutral hadrons in the calorimeter.

There were two presentations by the forward experiments, TOTEM and LHCf, showing that their apparatus is working and they are able to understand the signals to some degree.

The most exciting note was struck by the ALICE and CMS Collaborations, who gave a glimpse of their fast-track physics analyses. ALICE has already submitted a publication to EPJ on the measurement of the charged particle pseudorapidity (η) density (arXiv:0911.5430). They will also measure the anti-proton/proton ratio, which is sensitive to the physics of hadronization, and the p_T spectra of pions, kaons and protons. The CMS Collaboration will update the ALICE measurement of dN/dη, covering a wider range in η with much higher precision.

Jim Virdee, spokesman of CMS, delivering the CMS report. People were standing in the back and sitting in the aisles.

Clearly a tremendous amount of rapid work has been done by all collaborations, and the very brief presentations today showed only a very limited set of highlights. It is amazing how well the reconstruction and analysis of the data has proceeded already, and how well the simulations reproduce the data, especially at the level of the detector simulation. Of course this is only the beginning, but I can confidently predict and explosion of physics results to be published next year!

CERN Seminar: “LHC, Week 1″

A first public seminar at CERN – standing room only – was held today (Thanksgiving Day, 26-Nov-2009). The slides from the talks are available here: INDICO web page.

Steve Myers (CERN) kicked off the meeting with a pithy contrast of photos of a severely damaged set of magnets with a beautiful machine monitor trace showing manifestly stable beam. The audience responded with enthusiastic applause. The progress leading to the first collisions was amazing, as he told it. For example, they were able to obtain the “beta beat” of the machine on the first try – it took five years to obtain this with LEP. The bottom line: they circulated both beams 2-1/2 days after starting to circulate the first beam, and all four experiments recorded collisions some hours after that (p.40 of his talk). As Myers points out, thorough preparations pay off. The cryo system – the largest in the world by far – has worked flawlessly since Oct 8. The new magnet system which failed catastrophically due a splice resistance of 220 nano-Ohms last year, has no magnet with a resistance above 1 nano-Ohm today. And they are all protected by the new quench protection system.

Frederico Antinori (INFN Padova) presented the results from the ALICE Collaboration. The first event display showing a collision appeared seconds after colliding beams were present, and they recorded some hundreds of such events with typically 20 charged tracks. He showed good timing of their beam scintillators, and a beautiful vertex distribution obtained online from their higher-level trigger monitor some minutes into the run! The vertex distribution reflecting the luminous region, is 475 um in transverse direction and 4.2 cm in the longitudinal direction. This is quite impressive, showing that the tracker works well, that it is aligned, and that their software is ready. Jet analyses should follow soon. (He finished with a clip of the goings on in the ALICE control room when the first collisions were observed, but for some reason his slides are not linked at the INDICO web site.)

Andreas Hoecker (CERN) presented the results from the ATLAS Collaboration. All subsystems were operational, but a few were off for safety reasons. Interestingly, their muon toroid field was on although their solenoid field was necessarily off. They were able to record all beam splash events, and their forward tagging performed well. They could check the accuracy of their timing with the splash events. There is a beautiful event display showing the bending of beam halo events, and of course the first collision event (45 tracks!). They have good confidence again based on timing measurements – from their liquid argon calorimeter, good to 1.5 ns, among other measures. The “cogging” of the beam gives a shift of the impact parameters from good events exactly as expected. They see about 9 GeV of calorimeter energy, consistent with zero missing energy, well reproduced by their simulation (rms 1.2 GeV on the projection). It is exciting to see a di-jet candidate, with the jets in the forward direction and transverse energy of roughly 10 GeV. From 197 golden candidate events, they derive a very rough estimate of 4.9 mb^-1 integrated luminosity.

Ivan Mikulec (HEPHY – Vienna) presented the results from the CMS Collaboration (of which I am a member). Beam splash events verified that the timing of the detector is greatly improved compared to last year, for example in the HCAL. Beam halo events left clear tracks in the cathode strip chambers (as I described a few days ago). Ivan showed trigger rates during the now famous “Monday afternoon” fill. Reconstruction of primary vertices shows a narrow peak with and rms of 4.6 cm. Energy losses (dE/dX) is consistent with min-ionizing tracks. Hits in the ECAL and HCAL calorimeter give information on the timing, consistent with collisions. The grand finale was a reconstructed pi-zero peak. The best peak has a width of 10 MeV, and a peak position a little below the nominal pi-zero mass, due to the fact that the magnetic field is not on (the effect is predicted by the simulation).

Olivier Callot (LAL-Orsay), a former ALEPH colleague, reported on behalf of LHCb. On page 6 he showed a beautiful display indicating the time-evolution of the beam splash events (animation on page 7). LHCb also observed beam-gas events, which are rather important for them right now, and confirmed them on the basis of the beam crossing number. Very pretty track were reconstructed in their “velo” (bicycle) tracker. LHCb also sees a very nice pi-zero peak, with the correct mass. Their Ring-Imaging Cerenkov Detector recorded some beautiful rings from beam-induced interactions. Collision events give a higher sum of transverse energy than beam gas events, and nice vertices can be reconstructed. The Z distribution of the vertices shows a beautiful peak a the correct position, with a width of 10cm.

Olivier’s closing remarks are the perfect summary:

This machine is fantastic!

and, of course, all collaborations are ready for more…

A final observation: Tommaso Dorigo, a famous physics blogger, was present in the front of the auditorium:

Audience in the CERN Auditorium while Steve Myers speaks.

Examining Collision Events

Inside CMS, people are busy learning what they can from the handful of collision events recorded yesterday. Our magnet was not on at the time because it would have interfered with the beams, at this early stage, so we cannot understand anything about charged track momenta. But those straight-line tracks still provide a testing ground for several key aspects of event reconstruction.

A close-up of straight-line tracks from the earliest LHC collisions

Since I am a loyal member of the CMS Collaboration, I won’t divulge here any details about what my colleagues are doing, but I will say that I am very impressed by what I see. Stay tuned – public presentations are scheduled in a matter of days…

Collisions in CMS !!!

CMS and the other LHC experiments have seen the first LHC collisions ever!

One of the very first CMS Collision Events at 900 GeV

You can clearly see a slew of green tracks coming from the interaction point. They are all straight because the magnetic field is off today. (It will be turned on later – remember that the experimental solenoids affect the beam so the LHC operators have do carry out careful measurements to compensate for these effects – to be done asap.) You can also see energy registered in the calorimetry (red is the hadronic calorimeter, and blue is the electromagnetic calorimeter). There are no muons in this event, and muons are not expected in this kind of minimum bias event. (They may appear as a consequence of the decays of pions and kaons – more about that later.)

I need to board my flight now (I am en route back to my home institution) – see Darin Acosta’s log for good, real-time reporting of what his happing in the CMS experiment.

This is a wonderful, wonderful day!

Can we turn on the High Voltage?

If you want to turn on a light, or start your car, you rarely pause to think about possible damage that might result. But when beam is coursing through the CMS muon end caps, we think about it very carefully. In fact, we discuss in all seriousness when to turn on the high voltage, when there is beam in the machine.

Images of the CMS muon end cap detector (EMU for short) have been shown hundreds of times, including on the cover of Newsweek. Lots of copper surfaces, thick steel disks – what could be delicate about that?

One of the CMS muon end cap disks being lowered into the experimental hall

The chambers inside are very well made by experts in Russia, China and the United States. They are relatively robust in a mechanical sense, though we would not want any of them to fall from their mounting on those large steel disks.

The danger is in the gas and the wires. The detecting layers consist of thin gas layers sandwiched between cathode strips, with anode wires stretched in planes between the cathode planes. A large voltage (several thousand volts) are applied between the cathodes and the anode; this is part of the gas amplification mechanism which allows us to detector the wispy muon tracks as they pass through the chamber. In order to detect the muons, we need to have this high voltage turned on.

Diagram of the gas gap in a cathode strip chamber

This is a delicate instrument, and quite sensitive to minute amounts of ionization (on the order of 100 electron-ion pairs or fewer). You can’t spray it with intense concentrated charged particle fluxes or you risk damaging the instrument. Putting the beam through one of these chambers would be like igniting an old-fashioned camera flash in front of night vision goggles.

You might object that we splashed millions of muons through the detector in the famous beam splash events, which I wrote about a couple of days ago. It is true that this is an immense amount of ionization compared to a single muon or a muon pair. But it was spread around a hundred squared-meters of area, not concentrated in a narrow area close to the beam. (Nonetheless, we put the high voltage at low values just to be very safe, as mentioned in my posting.)

So, when the LHC operators are circulating beam, looking for the rough spots and making systematic adjustments to machine controls and monitors, we have to protect our delicate detector. At this point in time, when an eager physicist asks:

Can we turn on the high voltage?

the answer will often be:

be patient, not quite yet…

Beyond Beam Splash Events

Yesterday we enjoyed our second set of beam splash events, generated with beam one in contrast to earlier splash events generated with beam two.

Today we are thrilled to see beautiful beam halo events in CMS, like this one:

Event display of a nice beam halo event from CMS

You can clearly see a trajectory (in red) extending across the CMS detector based on short track segments (light blue dashes) reconstructed in the muon endcap systems. (The grey cylinder in the middle indicates the tracking volume.) This picture was made by my graduate student as part of our effort to understand and validate the performance of the CMS cathode strip chambers (which are traced here in dark blue).

Muons in beam halo events run parallel to the beam and typically have a few hundred GeV, according to simulations. (We would love to test that with the real data…) They are generated when protons from the beam pass out of the beam pipe and strike some object near by, leading to an energetic hadronic shower out of which emerge one or more muons. This shower occurs many tens or hundreds of meters away from the experiment, so any muons that reach the apparatus have hardly any angle with respect to the beam.

These beam halo muons may be a nuisance one day, but right now they are a novelty. In fact, they are quite useful for the end cap muon systems, since they provide straight lines through the muon chambers which can be used to refine the chamber alignment, and their arrival time is well-defined thanks to the bunched nature of the LHC proton beams. (For some nice illustrations, see the postings by Darin Acosta.) This strobe-like signal helps us refine the synchronization of the chamber signals, which is important for recording them properly and for defining an accurate and efficiency trigger.

Two months ago we were busy analyzing cosmic rays. Two weeks ago, and two days ago, we were extracting information from beam splash events. Today, and for the next few days, we are looking at beam halo tracks. What comes after that? collisions!.

Beam Splash Events in the CMS Muon End Caps

Excitement returns to CMS this month, as the LHC begins to circulate beam. There are many good sources of information, for example, the online commentary by Darin Acosta, among others.

My team from Northwestern University is busy providing prompt feedback on the response of the cathode strip chambers (CSCs) from the CMS experiment. On 9-November, we observed the beam splash events produced when Beam 2 struck collimators and a wall of muons passed from the -Z to the +Z side of CMS. Here is a depiction of the charge measured on the radial strips of the CSCs:

Beam 2 splash event in the CSCs

The arrow indicates the direction of Beam 2, and one sees clearly more charge on the strips on the upstream side compared to the downstream side. The red fans show the inner set of chambers, while the blue fans show the outer. (There is one pair of green fans, but they are too small and to faint to make out in this picture.)

Here is a new event from this evening, 20-November, in which Beam 1 produces a splash in the CSCs:

Beam 1 splash event in the CSCs

Comparing to the picture above, it is clear that the two muon endcaps have exchanged rolls (and indeed, Andy reversed the direction of the arrow).

It is worth noting that the HV is set to stand-by values. The flux of muons is so great, on the order of 5 muons per cm², that we nonetheless see a tremendous about of charge compared to what we expect for a normal single muon, such as a cosmic ray or one coming from a pp collision.

A more conventional, and colorful, view of these kinds of events is given by the official iSpy event display program. Here is an example:

Beam Splash event as depicted by iSpy

The purple parts in the end caps are the CSCs, obviously registering lots of charge while many other subdetector systems are off.

As I write this post, the LHC operators are `capturing’ the beam, which means that the protons’ orbit is determined by the RF cavities that are turned on. This is a major milestone on the way to collisions.

A Search for Collinear Muons

The D0 Collaboration recently posted a brief paper describing a search for Higgs bosons in the NMSSM (arXiv:0905.3381). The model supposes that the scalar Higgs boson, h, decays predominantly to a pair of very light pseudo-scalar bosons, a. If the a bosons are very light, then they may be expected to decay predominantly to a pair of muons or tau leptons. So the signature would include events with four muons (very distinctive!) or two muons and the decay products of two tau leptons (also distinctive).

I like the idea of finding a light particle produced in high-energy collisions through its decay to muons (or photons or electrons) since it helps underscore the fact that hadron collider experiments are also a good place to look for very rare processes, not just heavy particles. I wrote earlier this year about looking for a bosons at a hadron collider and it is very nice to see this analysis by D0.

The physicists who conducted this analysis were faced with at least one interesting challenge: reconstructing two high-energy muons which come close together in space. (The muons are close together because they come, hypothetically, from a fairly light particle which itself comes from the decay of a fairly heavy particle. So the muons have a large Lorentz boost in the laboratory frame.) Most muon detectors are designed to register a single muon well, or perhaps two muons that land far apart. Their granularity is poor, compared to tracking devices or even the calorimetry. There can be problems with producing a valid muon trigger, and also with reconstructing the muon tracks themselves, even offline. Finally, one has to be careful when demanding that the muons be isolated, since they are not isolated, strictly speaking.

For the four-muon channel, the D0 physicists approach this challenge by asking only that one muon out of each pair be reconstructed, and then they pair each of two reconstructed muons with a “companion” track, meant to be the muons that were not reconstructed successfully. This may sound like it should lead to a large background, but remarkably, it does not, thanks to the isolation criteria they applied. Only two events are selected in over 4 fb^-1, consistent with expected backgrounds, and neither of these has more than two muons.

The reconstruction of nearly-collinear tau decays is much harder. In fact, the D0 group did not try to reconstruct the tau’s directly, but rather leaned on the fairly large missing energy coming from the neutrinos emitted in the tau decays. (For a nice simulated event, see the analysis web page.) With two muons, significant missing energy, and then evidence of tau decays and vetos against jets, again the expected background is quite small. The two muons from the a decay (remember, for this channel one a decays to a muon pair, and the other to a tau pair) can be used to reconstruct the mass of the a boson. So the natural strategy is to look for a bump in the di-muon spectrum for this sample of events.

Here is the D0 result:

di-muon spectrum for candidate events with two muons and two tau leptons

The eye catches the “peak” at 4 GeV, but of course this peak is not statistically significant. It is nice that the D0 physicists have produced a smooth background prediction from data (blue dashed histogram) – the selected events appear quite consistent with that. Narrow peaks are also drawn on the graph, indicating hypothetical signals, as black curves and black histograms. These are meant for illustration, to show the narrowness of the peak that might have been reconstructed
with the D0 detector, if a real signal had been present.

The D0 paper contains some limits placed on this production channel, which require a few not unreasonable assumptions. I’m not so interested in those limits, which generally are higher than the theory predicts. What pleases me about this piece of work is the ability of the D0 detector and event reconstruction to look for such events.