[I extorted the text below from my friend and colleague Andrea Giammanco, who has been one of the driving forces behind a new scientific result at the crossroads of nuclear and particle physics - TD]
A new result by the CMS collaboration [CMS-PAS-HIN-19-001, cds.cern.ch/record/2699428] demonstrates for the first time that top quarks can be produced in nucleus-nucleus collisions.
To provide the context for this breakthrough, let us start by recalling that the flagship accelerator of CERN is named Large Hadron Collider (LHC) and not Large Proton Collider, in spite of being most famous for how its record-energy proton-proton collisions brought the discovery of the Higgs boson and a vast program of searches for new physics.
In a typical year of LHC operations, in fact, while six months are spent for proton-proton collisions at the highest achievable centre-of-mass energy, one is devoted to heavy ion beams. The two kind of runs are for very different kinds of physics: proton-proton collisions are, for example, how we precisely pinned down the properties of top quark, while heavy ion collisions are used to investigate dense nuclear matter at extreme energy. In particular, lead-lead collisions at the LHC have already clarified several properties of the quark-gluon plasma (QGP), an extreme state of matter that filled the Universe in its first microseconds of life, but also opened new questions about it.
The top quark, being the heaviest elementary particle known (see figure below), is also a unique and potentially very powerful probe of the inner content of nuclear matter. Let us see why.
[Caption: Masses of the massive vector and scalar bosons (green squares), quarks (orange circles) and charged leptons (blue diamonds). This figure includes all the fundamental particles whose mass is non-zero and has been measured. Image credit: Pieter David.]
First, because to create a pair (particle and antiparticle) of such a massive quark one needs to convert a large amount of kinetic energy into mass through a single elementary interaction. As the "partons" (quarks and gluons) inside "nucleons" (protons and neutrons) have a very large spread in the fraction of nucleon momentum that they carry, only a tiny fraction of the parton collisions pass the threshold for top quark production. This means that by selecting events containing top quarks one is implicitly studying only the most energetic parton-level collisions, which are the most poorly known, especially when they involve gluons (and gluons are involved in the vast majority of the collisions leading to top quark pair production.)
Second, because the top quark's mass causes it to decay extremely fast. That's because it is heavier than the W boson and this fact, unique among all quarks, means that its main decay channel, mediated by the weak force, is a 2-body decay into a real W boson (that further decays to leptons or quarks) and - usually - a bottom quark. That makes a substantial difference in decay speed with respect to all other unstable quarks, which decay through a 3-body branching that involves a "virtual" W boson, where the large W mass acts as a wall (being heavier than the quark in the initial state) that can only be punched through via quantum tunneling. For example, while the average lifetime of a top quark is of the order of a yoctosecond (10-24 s), the lifetime of its sibling the bottom quark is of the order of a picosecond (10-12 s).
Why does it matter? Because a yoctosecond is 2-3 orders of magnitude smaller than the typical processes mediated by quantum chromo-dynamics (QCD) that are unescapable by the other quarks. This includes hadronisation (the process that "dresses" all other partons until the final state only contains particles that are neutral in terms of "color", which is the QCD equivalent of charge) and the formation of the QGP, whose expansion and cooling are roughly estimated, from theoretical consideration, to take order of 10-22 seconds.
For the reasons above, one expects that top quarks can provide insights on the highest-energy parton collisions, complementary to the studies already performed with other "probes" such as other quarks and the electroweak bosons , W and Z. In particular, we eventually intend to use its feature of decaying within the QGP into other quarks, which, differently from the top, experience intense interaction with the QGP, allowing a study of how the latter "quenches" their energy.
However, that's not so easy. Heavy ion collisions produce a humongous number of particles, putting all the algorithms of CMS under stress, with some of them needing a dedicated re-optimization in order to work at all. Indeed CMS (as well as ATLAS) is a multipurpose experiment, mostly optimized for high-energy proton-proton collisions; therefore, it would be unfair to expect it to be equally performant in heavy ion data, which are the realm of ALICE, a dedicated experiment at the LHC. On the other hand, ALICE was never expected to be able to select top quarks.
It took some years to build confidence that this was possible at all, passing through some intermediate steps.
The top and heavy ion experts of the CMS collaboration started collaborating for the first time with the analysis of a very short proton-proton run (just two days carved out of the busy LHC schedule) taken in 2015 at a centre-of-mass energy of 5.02 TeV. This number may appear odd, but it is what you need in order to compare proton-proton data to lead-lead ones, as it is the energy carried by a single nucleon inside a lead ion accelerated to the largest achievable energy (roughly given by the maximum proton beam energy times the number of protons Z, and distributed across A nucleons, with Z and A being respectively 82 and 208 for the lead isotope employed in the LHC).
The first measurement of the top quark production cross section at that particular energy [https://arxiv.org/abs/1711.03143] was already motivated by the need of a reference point for future top quark studies in lead-lead data, but it was also used to set a new constraint on the parton distributions within the proton, as its modest statistical precision was compensated by the energy being intermediate between Tevatron and the Run 1 of the LHC (which took data at 7 and 8 TeV between 2010 and 2012).
The next step, and already a breakthrough, was the first observation of top quarks in proton-lead collisions using data taken in 2016 [https://arxiv.org/abs/1709.07411]. Such collisions are mostly motivated by the need to deconvolute QGP effects from other properties of nuclear matter at high energy, such as modifications in the gluon distribution within a nucleon that are known to depend on the size of the nucleus. Again, although this first measurement of top quark production cross section in proton-nucleus collisions was originally motivated as a milestone towards future lead-lead data (it was the baptism of fire for top quark selections in a nuclear environment), it stood on its legs as an important input to global fits that seek to constrain the aforementioned gluon distributions within nucleons bound in nuclei.
In spite of the experience acquired with proton-lead data, the problem of top quark identification in lead-lead collisions was still a tougher nut to crack.
[Caption: a lead-lead collision event interpreted as originating from the chain gg--> tt -->Wb Wb --> eν μν bb.]
The figure above shows an actual example of how a top quark event appears, in the busy environment created by a heads-on lead-lead collision inside CMS. The number of background particles is overwhelming, but some features of the top quark signal are recognizable: a very energetic electron leaving a large deposit in the electromagnetic calorimeter, a very energetic muon reaching the end of the dedicated muon detectors, and two very energetic hadronic jets containing signs (not visible in this picture) of originating from the hadronization of b quarks (hence passing a so called "b tagging" algorithm, which had to be retuned specifically for this lead-lead run in order not to be almost blinded by this large particle multiplicity). We interpret this event as originating from the chain gg --> tt --> Wb Wb --> eν μν bb, with the two neutrinos gone undetected.
Let us spend a few words about those b-tagged jets, and which role we want them to play. Being present in almost 100% of the top quark decays, b quarks are a crucial signature of top quark presence. On the other hand, as mentioned above, b quarks are not oblivious to being surrounded by a QGP, and this creates a "circular logic" concern (and therefore a potential bias) if we then intend to interpret the results to understand the QGP itself.
Two complementary analyses were performed: one based entirely on the features of the two leptons observed in the event, intentionally ignoring any information related to the hadronic jets in order to stay away from such subtleties; and another analysis where the presence of b-tagged jets is also exploited, as they are a tell-tale sign of the presence of top quarks in the events, with extra uncertainties added as nuisance parameters in the final fit to the data. In both cases two light charged leptons (electron or muon) are selected, with opposite charge sign. And in both cases a Machine Learning algorithm (a boosted decision tree, or BDT) is used to combine the most relevant properties of the two-lepton system (e.g., how their directions differ from being back-to-back in the transverse plane, as back-to-back leptons are what one expects from the most likely source of background, which is the so called Drell-Yan production).
The output of the BDT, for different b-tagging multiplicities, is compared to the best-fit signal+background expectation in the electron+muon channel in the next figure.
[Caption: data and expectations in bins of the BDT discriminator and b-tagged jet multiplicities, for electron+muon events. Expectations for signal and background are scaled to the best-fit values.]
The two analyses yield results consistent with each other and with extrapolations from previous measurements of top-antitop cross section in proton-proton collisions at the same centre-of-mass energy per nucleon.
[Caption: results of the two alternative analyses (with and without exploiting the b-tagged jets), scaled by (208)^2 in order to compare with the measurement in proton-proton collisions at the same nucleon-nucleon centre-of-mass energy, and also compared to theory expectations.]
Going back to the original motivation, a big question in heavy ion physics is still open: how much time does the QGP take to form, and then to cool down? Recent theory literature [https://arxiv.org/abs/1711.03105] elucidated how to use the different steps in top quark's production and decay as a sort of yoctosecond-scale chronometer, uniquely fit to answer that question. More data will be needed to get there (and luckily, a tenfold increase in the size of the lead-lead dataset is expected during the next LHC runs), but this very first step has been crucial, as it demonstrated for the first time that identifying the presence of top quarks in such a messy collision environment is feasible.
