Banner
    CMS Sees Hint Of Upsilon Suppression In Quark-Gluon Plasma!
    By Tommaso Dorigo | May 23rd 2011 07:22 AM | 22 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
    A few months ago LHC took a special run of proton-proton collisions at  2.76 TeV. Why the lower energy, now that we are accustomed to searching for new phenomena at the highest available energy of 7 TeV ? Because of the wish to compare lead-lead collisions, taken last year at 2.76 TeV nucleon-nucleon energy, with proton-proton ones. The comparison allows to extract extremely interesting results.

    When heavy nuclei collide at very high energy, they are thought to create, for a very short time, a hot plasma of hadronic matter which is unlike anything else in the Universe today. Proton-proton collisions do not do that, because there are too few energetic partons around to create a significant "blob". The comparison between what we observe in heavy ion collisions and proton-proton collisions taken at the same equivalent energy allows us to draw some interesting conclusions on the behaviour of this plasma.

    In a hot plasma of quarks and gluons, it should be harder for stable hadrons to retain their binding. So one specific signature of the creation of this extreme condition of matter is the fact thaw we should detect a reduced yield of heavy mesons -bound states of heavy quark-antiquark pairs which are not the typical product of normal parton fragmentation processes, but which are indeed produced by the hard collision. Ok, but what kind of mesons should we go after ?

    Of course, we need to focus on mesons whose creation occurs at the instant of the hard collision, and whose presence we detect with great precision: ones which are unmistakable even in the horrendous mess created by a collision of hundreds of nuclei in simultaneity. This makes the Upsilon meson, a bound state of a bottom-antibottom quark pair, an ideal testing ground: the Upsilon and its excitations (labeled by their angular momentum: 1S, 2S, 3S... in ancient spectroscopic notation) decay to muon pairs, which escape unharmed the hot soup, and get detected with precision in the outer detecting elements of CMS.

    Where Have All The Upsilons Gone ?

    Have a look at how we see Upsilon particles below: the three resonances 1S, 2S, and 3S stand out of backgrounds very cleanly in proton-proton collision data. Their mass is reconstructed easily from the measured momentum of the escaping muons.



    And now look at the same graph obtained from the lead-lead collisions. Is there something strange going on here ? Indeed the 2S and 3S states appear to have vanished -or at least to be strongly reduced from the yield they showed in proton-proton data. The ratio between production of 1S and (2S plus 3S) states should be the same in the two datasets, because the production of these hadrons is governed by the same mechanisms: a nucleon is a nucleon is a nucleon for the strong force... But it apparently isn't!


    (Two words of explanation of the graphs: the black dots with error bars show the dimuon invariant mass in the data -respectively proton-proton collisions in the top figure and lead-lead collisions in the bottom figure; the blue curves show the result of unbinned likelihood fits to the mass spectra, using some fixed parameters to aid the fit, such as the mass difference of the three states fixed to world average value. The red hatched curve in the bottom figure reports the expected yield of signal in lead-lead collision data from the yield observed in proton-proton collisions. The background is allowed to vary shape parameters in each fit).

    The suppression of the excitated Y states is predicted by the "melting" of these bound states in the quark-gluon plasma. Since the 2S and 3S resonances are less tightly bound than the 1S, they melt more readily in the high-temperature environment, and we see fewer of them. This, at least, is the suggestion of the analysis, which finds a 2.4-sigma deficit of 2S and 3S candidates with respect to 1S candidates in lead-lead collisions, taking the relative yield in proton-proton collisions as a reference.

    2.4 standard deviations from a unit ratio is not a huge effect, and the possibility that we are just hyping a statistical fluctuation does exist. However, I tend to believe that this signal is genuine. Time will tell, of course, when more heavy ion collisions will be recorded by the LHC experiments. Also, ATLAS and ALICE will also produce results on this effect, I am sure.

    An interesting note - for experts only

    Now, for experts I would like to add something at the bottom of this post. What experiments claim in observations of this kind is a "suppression of particle production". They mean that the bottom-antibottom resonances are not forming in the hot gluon plasma, because of the dense environment and its complicated dynamics. I question the validity of this assertion, both from a theoretical and from an experimental point of view.

    Theoretically, what we might envision is that the bottom-antibottom pair gets created and does bind; the resonance then has a relatively long lifetime because quantum chromodynamics can only disintegrate it by letting the pair emit no less than three gluons (the Okubo-Zweig rule). In short, one gluon cannot mediate the disintegration because the Y is colourless; two gluons have the wrong C-parity, which is a conserved quantity in strong interactions; so three gluons are the minimum. The simultaneous emission of three gluons is thus an alpha_s^3 effect, and is thus suppressed enough that the competing decay to dimuon pairs through an electroweak annihilation diagram emerges a significant fraction of the time (a few percent). Thanks to all this, we do detect the Y decay!

    Now, the point. In a hot gluon plasma, the resonance can exchange degrees of freedom more readily, and the conservation of colour or C-conjugation quantum numbers may be taken care of by the plasma itself. If the strong decay then proceeds more readily, the electroweak process is suppressed.

    Experimentally, we count only dimuon decays -we do not have any access to measuring the hadronic branching fraction!-, so we observe a deficit. That deficit is the result of the suppression of "production cross section times branching fraction into dimuon pairs", and not just suppression of the production!

    I think this is a subtle point which I have however never heard disproven in a conclusive way. I would be very, very happy if some of you knowledgeable readers could shed some light for me into the (hot) matter. Thanks!

    Further reading: see here.

    Comments

    Vladimir Kalitvianski
    I am not good at this subject and I did not do any estimations, hence my question: why they call it a quark-gluon plasma? If I take a couple of very fast atoms (systems with bound electrons) and collide them, they do not form any plasma whatever the number of electrons is in them. In a hot plasma you have a distribution of particles in the velocity (momentum) space, for example, a Maxwell distribution. It is a rather smooth distribution. Two fast atoms form quite different initial and final distributions, the distributions are rather narrow, kind of delta-functions of velocities. So it is not a hot plasma but a very weakly interacting systems. The Born approximation is more valid then. Even large angle scattering with full ionization (breaking binds) is a sudden perturbation without having time to get everything averaged, put in equilibrium, etc. So it is not a plasma at all.

    Similarly, in lead-lead nuclear collisions there is no plasma but "frozen" weakly interacting systems. Of partons, if you like.
    dorigo
    Hi Vladimir,

    I think there are good reasons to call it a plasma. The temperature is extreme, and although the state exists for a very short time, it has all the characteristics of a thermal gluon bath. Please do not confuse atoms with nuclei: quantum electrodynamics is different from quantum chromodynamics...

    Cheers,
    T.
    Vladimir Kalitvianski
    Yes, they are different but what do estimations say? How long does this "plasma" exist? Does it have time to get thermalized? Aren't partons "frozen" during collision? Is there a "temperature" in a sense of equilibrium? Or just large relative energies are sufficient to call it a "hot plasma".
    dorigo
    Suppression of many things has been seen in QGP phenomena. Jet quenching was seen by CMS recently, too. However, this to my knowledge is the first clear hint of Y suppression.

    Best,
    T.
    Bonny Bonobo alias Brat
    The suppression of the excitated Y states is predicted by the "melting" of these bound states in the quark-gluon plasma. Since the 2S and 3S resonances are less tightly bound than the 1S, they melt more readily in the high-temperature environment, and we see fewer of them. This, at least, is the suggestion of the analysis, which finds a 2.4-sigma deficit of 2S and 3S candidates with respect to 1S candidates in lead-lead collisions, taking the relative yield in proton-proton collisions as a reference. 
    2.4 standard deviations from a unit ratio is not a huge effect, and the possibility that we are just hyping a statistical fluctuation does exist. However, I tend to believe that this signal is genuine. Time will tell, of course, when more heavy ion collisions will be recorded by the LHC experiments.
    Tommaso, can we be sure that no mini black holes are being created here? As mentioned in the arxiv paper referenced by kneemo above, jet suppression has also been seen at RHIC and these string theorists modelled the phase where meson melting occurred as a higher dimensional black hole. They said that :-
    Meson melting is a first order phase transition from D7 brane probe fluctuations with boundary conditions leading to infinitely long lived bound states (the mesons) to fluctuations of D7 branes that end on a black hole, where there the boundary condition gives states that decay by falling into the black hole (the quasinormal" modes - the melted mesons).
    Minkowski embeddings are related to the phase where we have mesons (quark bound states) in our gluon plasma, whereas the black hole embeddings are related to the phase where the mesons have melted in the plasma (the finite lifetime quasinormal modes arising from in-falling boundary conditions at the horizon). The meson melting transition occurs near where the meson mass is zero, (the far right of figure 5) but not exactly. 
    To find the exact location (in terms of the quark mass), we need to calculate the free energy of the system and find where the jump (determined by minimizing the free energy) between the black hole and Minkowski embeddings occurs, inside the region of multivaluedness of the condensate vs. mass curves of figure 2(d).
    Obviously I am way out of my depth here so a simple reassurance would be great :)














    My latest forum article 'Australian Researchers Discover Potential Blue Green Algae Cause & Treatment of Motor Neuron Disease (MND)&(ALS)' Parkinsons's and Alzheimer's can be found at http://www.science20.com/forums/medicine
    dorigo
    Helen, I assure that no black hole was created in the process. At least, nobody claimed so :)
    Cheers,
    T.
    hi,

    there are few details also in the page you link so I cannot understand the full analysis.
    One thing that puzzles me is how you compare analysis with protons and ions.
    In particular the cuts should be different. How do they checked that it is not an artifact of the analysis
    that result in the quench of the 2S and 3S? In dense enviroment it easy to screw up variables
    (such as isolations). .

    D

    dorigo
    Hi D,

    of course the efficiency of muon detection has been studied in excruciating detail in HI collisions too, and it checks okay when compared with pp data. I think no isolation cut was applied. But anyway, the muons from 1S and 2S and 3S have very similar kinematics, so this being a ratio measurement, all the systematics you worry about cancel. Even a large difference in efficiency in the two datasets would result in practically no effect on the double ratio R2=(N(3S)+N(2S)/N(1S))_HI/(N(3S)+N(2S)/N(1S))_pp. But there is almost no difference, as I said.

    Anyway, CMS published earlier a study of resonances decaying to dimuons in HI data, see e.g. here. All seems good!

    Best,
    T.
    Hi Tommaso,

    interesting question. I don't know much about QCD bound states and plasmas, but here are my thoughts.

    Usually bound states are described by some sort of Potential and the corresponding Schroedinger equation. The quark gluon plasma will change the shape of the potential. This will most likely affect the higher lying states (2S, 3S) more than the lowest lying state (1S). Then it is relatively easy to understand why e.g. the 2S, 3S states are not formed anymore, while the 1S state is still there.

    On the other hand, lets assume, as you suggested, that the presence of the QCD plasma changes the branching fractions of Y by enhancing the Y decay modes into gluons. This could happen e.g. by absorbing a gluon from the plasma, and then decaying to two gluons. In this case, one would have to come up with an explanation why this only increases the gluon gluon branching ratios for the higher Y resonances.

    This is my idea why the production suppression picture might be favorable over the suppressed branching fraction explanation. Whether it has anything to do with nature, I don't know.

    Cheers

    Can someone explain please how is the temperature of the quark-gluon plasma estimated?

    Thanks
    E.

    Vladimir Kalitvianski
    Tommaso, please give us here the typical momentum of a parton in the Pb nucleus and the relative parton momentum in Pb-Pb collisions (I mean target-projectile relative momentum).
    dorigo
    Vladimir, I said 2.76 TeV per nucleon. I cannot say more, because partons within nucleons come with a distribution of momentum governed by the nucleon PDF. There is no "typical" value... We can instead say that the typical parton-parton Q^2 in the hard subprocess generating the Y resonances is of the order of 150 GeV^2, since Y mesons are known to have a very soft momentum spectrum, so most of the Q^2 actually goes in the creation of the particle mass.

    Cheers,
    T.
    Vladimir Kalitvianski
     2.76 TeV per nucleon means roughly 0.9 TeV per parton.

    If the typical momentum of a parton in a nucleus is $ M_n /c \approx 0.9$
    GeV/c, the ratio is about $10^3$. So the partons are not moving (are not orbiting their usual orbits) in course of collision due to lack of time, right? Anyway, the typical confinement energy is much smaller than the collision energy. In that case the partons are clouds of a cold gas rather than a hot plasma.
    dorigo
    Vladimir, if things were so simple, I would switch to some other science.

    Anyway I haven't a clue where you get the 0.8 TeV per parton -there aren't 3.5 partons per nucleon, and their distribution in momentum is not a Dirac delta.

    Cheers,
    T.
    Vladimir Kalitvianski
    You know, I am not good at estimations, that's why I first asked you to give them.

    I find it strange when I ask somebody to clarify/justify his statements and receive nothing in response.
    dorigo
    You receive nothing because you ask questions that have no answer.
    There is no "typical" momentum of a parton inside a proton...
    I find your attempt at "understanding" whether the plasma forms in a PbPb collision using the "typical" momentum of a parton in a nucleon so ill-founded that I am not worried much by not providing you with the right data. In any case, the answer is: parton distribution functions.

    Cheers,
    T.
    Vladimir Kalitvianski
    But there is a notion of an average, mean in any distribution. Is it difficult to specify? Again, even if I ascribe the maximum energy to one parton, it is still way inferior to the collision energy.

    OK, maybe you can say how many collisions  (on average again) the Pb partons have during a frontal Pb-Pb collision ?
    As far as I heard, the quark-gluon plasma is better described as liquid rather than plasma, which is more a historic term. Don't know if this helps you.

    Vladimir Kalitvianski
    The question is not here. I wonder if the Pb nucleus can be called parton plasma before collision and if there are sufficient number of parton-parton collision during a Pb-Pb frontal collision. If partons are asymptotically free at this collision energy, then it is difficult to call it a plasma. Well, who cares?
    Hi Vladimir,

    At the end you raise a great point. Indeed, identifying collective behavior is key. Several such effects were already observed, so we're comforable with the name "Plasma". I'm not an QGP expert, so a quick google search found
    http://www.physi.uni-heidelberg.de/~reygers/lectures/2011/qgp/qgp_lectur...
    which seems to answer your questions, perhaps in more detail than you'd like.

    As for the beginning: Tommaso already explained (probably appears below on this page) why the average is irrrelevant. One small addition, just to demonstrate how misleading a common sense application of math can be here: QCD is a divergent theory - there are infinitly many gluons with vanishingly small energies.

    cheers,
    Amnon

    dorigo
    Averages do not give you much in the case of a quite asymmetric distribution, and once you take into account the fact that triggers select a small subset of the more energetic collisions, your average says nothing about the data anymore if you do not model the trigger effect.

    And nope, I cannot say how many collisions the lead partons produce.

    Cheers,
    T.
    In solid-state physics the erergy levels of an impurity atom are strongly affected by the dielectric constant of the surrounding medium. Wouldn't a similar effect happen here so energy levels could be calculated by assuming plasma has an effective dielectric constant ?

    Thanks
    Paul