From HighEnergy HeavyIon Collisions to Quark Matter Episode

  • Slides: 44
Download presentation
From High-Energy Heavy-Ion Collisions to Quark Matter Episode II : The art of experimental

From High-Energy Heavy-Ion Collisions to Quark Matter Episode II : The art of experimental high-energy (heavy-ion) physics ? ? What you see is not what you get… ? ? Carlos Lourenço, CERN, July, 2010 1

Reminder: the question and the path towards the answer Goals: produce a QCD medium

Reminder: the question and the path towards the answer Goals: produce a QCD medium where colour is deconfined and study how the quarks and gluons interact in such a medium Method: 1) Collide high-energy heavy nuclei to create hot and dense strongly interacting matter, over extended volumes 2) Use certain “signals” to “probe” the properties of the created matter and prove, beyond reasonable doubt, that we have seen a “new state of matter” Problem: It is not easy to read Mother Nature’s book… especially when we want to establish signatures of “new physics” Extraordinary claims require extraordinary evidence 2

The art of experimental (high-energy heavy-ion) physics… 1) Many experimental issues are crucial to

The art of experimental (high-energy heavy-ion) physics… 1) Many experimental issues are crucial to properly understand the measurements and derive a correct physics interpretation: • • • Acceptances and phase space windows Efficiencies (of track reconstruction, vertexing, trigger, etc) Resolutions (of mass, momentum, energy, etc) Backgrounds, feed-down decays and other “expected sources” Data selection Monte Carlo adjustments, calibrations and smearing Luminosity and trigger conditions Evaluation of systematic uncertainties and several other issues. . . 2) “New physics” often appears as excesses or suppressions with respect to “normal baselines”, which must be very carefully established, on the basis of “reference” physical processes and “elementary” collision systems If we misunderstand these issues we can miss an important discovery… or claim the “discovery” of non-existent “new physics” 3

My name is , James (a charming Bond) The J/ is a bound state

My name is , James (a charming Bond) The J/ is a bound state of a charm quark and a charm antiquark (the 1 S state) It has a mass of 3. 1 Ge. V and decays, e. g. , into a pair of muons, + - CDF 2005 The CDF experiment sees a beautiful “resonance”, on the top of a flat dimuon continuum, far from the acceptance edges, with a very good dimuon mass resolution (20 Me) and an excellent “signal over background” ratio… An ideal measurement… published 31 years after the J/ discovery. It would be easy to “discover” a new particle if it would show up in your experiment like that; but at the time of the discovery, things are always less obvious… And the signals of QGP formation are more difficult to see than a “ 5 -sigma peak” 4

The first time the J/ was suppressed. . . Phys. Rev. D 8 (73)

The first time the J/ was suppressed. . . Phys. Rev. D 8 (73) 2016 ’s 70 Early 1 nb ? 1 pb Lederman was a very careful person. . . and was not in a hurry to get the Nobel prize p-U at 29. 5 Ge. V 5

The paper gives plenty of detailed information, including all the numerical values… We can

The paper gives plenty of detailed information, including all the numerical values… We can fit the data to the sum of an exponential continuum and a Gaussian “peak” It works quite well, with a “resonance” centered at ~3. 2 Ge. V with ~600 Me. V dimuon mass resolution 6

I see a resonance; you see a resonance; did they really miss it ?

I see a resonance; you see a resonance; did they really miss it ? Maybe they also saw it ! And they did not claim the discovery of a new particle because. . . …they did not know how to name it The Christ particle? The Hicks boson? The Limon? The Pope particle? Not an easy choice. . . Imagine me arriving at Rome’s airport : Customs’ officer : Purpose of your visit to Italy ? Answer : Giving a lecture on “Pope suppression with nuclear collisions” 7

Resolutions and acceptances distort the reality Leon Lederman et al. Phys. Rev. D 8

Resolutions and acceptances distort the reality Leon Lederman et al. Phys. Rev. D 8 (73) 2016 1 nb 1 pb They were convinced that the experiment had a better dimuon mass resolution: a “resonant structure” should show up as a narrower peak… And they thought that “the bump” could be an artifact caused by the acceptance edge… In the early 1970’s it was not very easy to simulate the acceptance and resolution of a detector And if you misunderstand such “details”, you miss the ticket to Stockholm p-U at 29. 5 Ge. V M (Ge. V/c 2) 8

The double discovery of charm, in 1974 J Ting et al. BNL 976 1

The double discovery of charm, in 1974 J Ting et al. BNL 976 1 l e b No Richter et al. SLAC The new particle got a “composite” name: J / ( in France it is known as “le Gypsy” ) 9

Acceptances ? HERA-B What is the peak at M ~ 1. 8 Ge. V

Acceptances ? HERA-B What is the peak at M ~ 1. 8 Ge. V ? A signal of D 0 → + - decays ? Not really… Just a signal that the acceptance changes significantly in this region Excellent dimuon mass resolution Acceptance is the probability that a particle is detectable by the experiment It depends on the kinematical values (rapidity, p. T, etc) and can be calculated by Monte Carlo simulation, reproducing the detector limitations and the analysis selection procedures E 866 10

The dimuon acceptances depend on the magnetic fields, on the thickness of the muon

The dimuon acceptances depend on the magnetic fields, on the thickness of the muon filter, on the distance between the target and the detectors, etc A (%) without field 0. 9 < M < 1. 1 Ge. V NA 60 Acceptance (%) correlation in p. T and ylab p. T (Ge. V/c) ylab A (%) with 2. 5 T field acceptance effect Monte-Carlo 11

Acceptance effects on the J/ nuclear dependence Shown in terms of a with sp-A

Acceptance effects on the J/ nuclear dependence Shown in terms of a with sp-A = s 0 × Aa (a=1 no absorption) DY 0. 92 E 772 (at x. F~0) 1992 a(J/ ) ~ 0. 92 α E 772 Why has the value of a changed from E 772 to E 866 ? x. F E 866 0. 95 E 866 (at x. F~0) 1997 a(J/ ) ~ 0. 95 Because the understanding of acceptances improved… 12

The value of a has a strong dependence on x. F and p. T

The value of a has a strong dependence on x. F and p. T The incomplete p. T coverage distorts the pattern vs. x. F The correction of the (correlated) acceptance is crucial E 866 α The problem was identified because the p. T coverage in E 866 was better than in E 772 using MC acceptance and dσ/dp. T consistent with data E 866 collected data with three magnet settings, each covering a different phase space window 13

Phase space windows Assume that the f has the same rapidity distribution in pp

Phase space windows Assume that the f has the same rapidity distribution in pp and p-Pb collisions while the w is “shifted” pp 400 Ge. V p-Pb 400 Ge. V A detector measuring dimuons in the window 3. 3 < y < 4. 2 sees the y 0 = 3. 37 f / w ratio increase from pp to p-Pb, concluding that a(f) = a(w) + 0. 04 Another detector, covering only backward rapidity, would “see” the opposite result: a decrease of the f / w ratio from pp to p-Pb collisions… Both would be wrong ! The result depends on the probed phase space window ! We can only correct for acceptances within the phase space window where we have data. Extrapolations to full phase space require assuming kinematical distributions that we cannot check: the “measurement” becomes model dependent. Experiments with a narrow phase space coverage must be extremely careful in formulating their results ! 14

Efficiencies Even in the phase space window well covered by the detector, sometimes a

Efficiencies Even in the phase space window well covered by the detector, sometimes a particle is produced but is not detected: maybe the trigger system missed it; or the tracks were not reconstructed; or the interaction vertex could not be identified; etc. The measurements must be corrected for these detection inefficiencies They might be measured in “special data samples”, or estimated by MC simulation using the same algorithms as used for the reconstruction and analysis of the data Efficiencies which depend on the centrality of the heavy-ion collisions are particularly dangerous: if not accurately corrected they may look like anomalies and mistaken for “new physics” “centrality” 15

Multiple scattering and dimuon mass resolution J/ NA 51 E 605 ’ pp @

Multiple scattering and dimuon mass resolution J/ NA 51 E 605 ’ pp @ 450 Ge. V s. M(J/ ) = 300 Me. V M (Ge. V/c 2) Muons are “identified” by absorbing all other charged particles in a “hadron absorber”. . . But the muons suffer multiple scattering and energy loss while traversing this “muon filter” 16

Standard way of measuring dimuons (NA 50, PHENIX, ALICE, etc) muon trigger and tracking

Standard way of measuring dimuons (NA 50, PHENIX, ALICE, etc) muon trigger and tracking beam iron wall magnetic field target hadron absorber Muon Other The muons suffer multiple scattering and energy loss in the hadron absorber 17

Overcoming multiple scattering with vertex tracking muon trigger and tracking vertex tracker dipole field

Overcoming multiple scattering with vertex tracking muon trigger and tracking vertex tracker dipole field iron wall magnetic field targets hadron absorber muon other Concept used in NA 60 and CMS : • The hadron absorber allows us to trigger on collisions that produce dimuons • The muons are tracked in the vertex tracker, before they suffer multiple scattering in the hadron absorber, and are matched to the tracks of the muon chambers • We can also see if the muons come from the collision vertex or not. . . 18

Muon track matching in CMS events / 20 Me. V Very good dimuon mass

Muon track matching in CMS events / 20 Me. V Very good dimuon mass resolution from the matching of the muons to the silicon tracks = 35 Me. V barrel + endcaps 19

Kinetic freeze-out p-nucleus at 400 Ge. V Muon track matching in NA 60 The

Kinetic freeze-out p-nucleus at 400 Ge. V Muon track matching in NA 60 The muon track matching significantly improves the dimuon mass resolution Chemical In-In @ 158 Ge. V A freeze-out NA 60 20

Data selection (quality is better than quantity) A small but clean event sample is

Data selection (quality is better than quantity) A small but clean event sample is better than a large but “dirty” one. And statistical errors are much easier to deal with than systematic uncertainties. No “outliers” remain after a proper data selection and “re-calibration” It is advisable to reject data of poor quality… but it is forbidden to reject data points that do not agree with your theory ! [All data points are equal… but some are more equal than others…] 21

Signals, backgrounds and “excesses” Suppose the expected signal is a small fraction (1%) of

Signals, backgrounds and “excesses” Suppose the expected signal is a small fraction (1%) of the estimated background and the total number of measured muon pairs is larger than their sum: Total = Background + Expected. Signal + Excess For instance: 1000 = 10 (expected signal) + 970 (estimated background) + 20 (unexpected source) Great ! What would you say? the signal is increased by a factor 3 → Big “excess” → New physics ! or the background was underestimated by 2% ? To properly study a signal, we must understand its backgrounds ! PHENIX NA 38/50 1% 2% 22

Know your reference ! p. QCD calculates partonic processes, like qq → qq, qg

Know your reference ! p. QCD calculates partonic processes, like qq → qq, qg → qg, gg → gg But our beams (and targets) are made of protons, neutrons, antiprotons. . . not of quarks and gluons ! The probability that we find quarks, anti-quarks or gluons inside a proton depends on their fractional momenta and on the “resolution” of our probe: f(x, Q 2) ? parton distribution functions, PDFs gluons sea quarks valence quarks 23

How do we know what the parton densities are ? Hard Scatter Calculation Parton

How do we know what the parton densities are ? Hard Scatter Calculation Parton Density Functions 5 experiments Cross Section Calculation Measurement e- DIS e- g g q Drell-Yan q l+ l 24

What means MRST, CTEQ 6 M, etc ? • Each class of experiments (DIS,

What means MRST, CTEQ 6 M, etc ? • Each class of experiments (DIS, Drell-Yan, etc) gets part of the story No single experiment measures the full picture of the proton • The results from each experiment go into a global fit Not all experiments agree – there is an art to “average” them together • Two main groups are experts in this art : → Martin, Roberts, Stirling and Thorne MRST → Coordinated Theoretical-Experimental project on QCD CTEQ 25

Jet production: data versus perturbative QCD calculations The data points seem to agree with

Jet production: data versus perturbative QCD calculations The data points seem to agree with the p. QCD calculation, over 11 orders of magnitude… Except if you look at the high ET tail. . . on a linear scale, as (data-theory) / theory ? ET (Ge. V) A clear indication of quark substructure (compositeness) ! Really ? ? ? 26

The high-ET jet excess got renormalized into a new reference New sets of Parton

The high-ET jet excess got renormalized into a new reference New sets of Parton Distribution Functions were calculated, including the CDF data The excess is gone! The quarks do not have substructure after all. . . Big-shot theorist Average theorist Spokesperson Level of belief that this was “new physics” Other faculty on experiment Graduate student’s advisor Postdoc Graduate student doing analysis 27

Keep your eyes wide open… It’s better without the covers, Sir ! at least

Keep your eyes wide open… It’s better without the covers, Sir ! at least for the photo. . . There is a big difference between looking… and seeing 28

Back to the Parton Distribution Functions Is a free proton the same as a

Back to the Parton Distribution Functions Is a free proton the same as a proton inside a nucleus? No! There are “nuclear effects” modifying the parton distribution functions. The probability of finding partons of given x changes when the proton is in a nucleus. The “EKS 98 model” provides the ratio between the PDFs in a proton of a nucleus of mass number A and in a free proton: EKS 98 for Pb Anti-shadowing “Shadowing” or “anti-shadowing”: decrease or increase of the parton’s density in the nucleus, in a certain kinematic range Shadowing 29

Is this an important effect ? This implies a ~20% higher charm production cross

Is this an important effect ? This implies a ~20% higher charm production cross section in Pb-Pb collisions at the SPS and a ~40% lower value at the LHC, as compared to a linear extrapolation from pp collisions. Remarks: For a given collision energy and a given mass produced, the values of x depend on the rapidity range where the measurement is made. If the pp and Pb-Pb collisions are collected at different energies, the corrections for the nuclear effects are particularly tricky. We cannot directly compare heavy-ion and pp data. Important in the analysis of the SPS J/ suppression data. 30

Systematic effects are difficult to control To verify the understanding of systematic effects, it

Systematic effects are difficult to control To verify the understanding of systematic effects, it is important to redo the measurements in different configurations, in terms of magnetic field polarity and magnitude, hadron absorber thickness, beam intensity and energy, etc. The acceptances, efficiencies, signal/background ratio, resolutions, etc. , will change; but the physics results, obtained after all the corrections are made, must remain the same (within statistical errors) Important analyses should always be independently made by two different groups within the experiment and with different choices of model dependent assumptions If after all checks you still have doubts about your exciting “new physics” results — you should always doubt everything, especially exciting results — you should make a new and vastly improved measurement 31

J/ suppression in nuclear collisions: a QGP signal In a medium with deconfined quarks

J/ suppression in nuclear collisions: a QGP signal In a medium with deconfined quarks and gluons, the QCD potential is screened and the heavy quarkonium states are “dissolved” into open charm mesons Lattice QQbar free energy Different heavy quarkonium states have different binding energies and, hence, are dissolved at successive thresholds in energy density or temperature of the medium; their suppression pattern works as a “thermometer” of the produced QCD matter T and similar for the Upsilon family 32

A “smoking gun” signature of QGP formation Feed-down from higher states leads to “step-wise”

A “smoking gun” signature of QGP formation Feed-down from higher states leads to “step-wise” J/ and suppression patterns. The suppression of bottomonium states is a very important component of the LHC HI physics program. J/ cocktail: ~ 8% from ’ decays ~ 25% from c decays ~ 67% direct J/ ’ c 33

The J/ and ’ normal nuclear absorption versus L The J/ and ’ production

The J/ and ’ normal nuclear absorption versus L The J/ and ’ production cross sections scale less than linearly with the number of target nucleons (while the high-mass Drell-Yan dimuons scale linearly). The Glauber formalism describes the J/ and ’ “normal nuclear absorption”, in p-nucleus collisions, in terms of the average path length, L, which they traverse in the target nucleus, from the c-cbar production point to the nuclear “surface” Projectile p-A 400 Ge. V J/ L Target ’ 34

J/ suppression in Pb-Pb collisions at the SPS p-Be p-A p-Pb central Pb-Pb e

J/ suppression in Pb-Pb collisions at the SPS p-Be p-A p-Pb central Pb-Pb e da c n e r refe S-U ta ess e proc c n e r refe Pb-Pb J/ normal nuclear absorption curve NA 38 / NA 51 / NA 50 The J/ yield (per Drell-Yan dimuon) is “slightly smaller” in p-Pb collisions than in p-Be collisions; and is strongly suppressed in central Pb-Pb collisions Drell-Yan dimuons are not affected by the dense medium they cross Interpretation: strongly bound c-cbar pairs (our probe) are “anomalously dissolved” by the QCD medium created in central Pb-Pb collisions at SPS energies 35

Measured / Expected In-In data vs. a sharp step function in Npart 1 A

Measured / Expected In-In data vs. a sharp step function in Npart 1 A 2 Step position Npart Step at Npart = 86 ± 8 A 1 = 0. 98 ± 0. 02 A 2 = 0. 84 ± 0. 01 2/ndf = 0. 75 (ndf = 5) Taking into account the EZDC resolution, the measured pattern is perfectly compatible with a step function in Npart 36

Measured / Expected What about the Pb-Pb pattern? Another step? 1 A 2 A

Measured / Expected What about the Pb-Pb pattern? Another step? 1 A 2 A 3 Step positions Npart Steps: Npart = 90 ± 5 and 247 ± 19 A 1 = 0. 96 ± 0. 02 ’ A 2 = 0. 84 ± 0. 01 c A 3 = 0. 63 ± 0. 03 2/ndf = 0. 72 Fitting the In-In and Pb-Pb data with one single step leads to 2/ndf = 5 ! In summary, the Pb-Pb pattern 1) rules out the single-step function and 2) indicates the existence of a second step. . . 37

J/ survival probability Back to the ideal world The predicted patterns, before any data

J/ survival probability Back to the ideal world The predicted patterns, before any data points were available, are quite different from each other It must be very easy to discriminate between the two! Right? normal nuclear absorption suppression by QGP Energy density We need measurements, to rule out one of these two scenarios (or both) 38

Let’s consider some recent observations made at CERN Can any of the models describe

Let’s consider some recent observations made at CERN Can any of the models describe all the experimental data points? 39

Data versus the “no new physics” model normal nuclear absorption “outlier” point; to be

Data versus the “no new physics” model normal nuclear absorption “outlier” point; to be rejected All kept data points agree with the expected normal nuclear absorption pattern! 40

Data versus the “new physics” model calibration error anomalous suppression All kept data points

Data versus the “new physics” model calibration error anomalous suppression All kept data points agree with the expected QGP suppression pattern! 41

Take-home messages 1) There is a BIG difference between “the measurements are compatible with

Take-home messages 1) There is a BIG difference between “the measurements are compatible with the model expectations. . . ” and “the measurements show beyond reasonable doubt that the model is good” 2) Nature never tells you when you are right, only when you are wrong. . . [Bacon, Popper] 42

How to discover your own “anomaly”. . . Playing with acceptances, efficiencies, backgrounds and

How to discover your own “anomaly”. . . Playing with acceptances, efficiencies, backgrounds and “well-known” references, you can easily find “anomalies” in your data; but remember that people around you will be much more careful and you may end up looking silly what’s wrong? why are you all looking at me? 43/47

More take-home messages Experimental studies of high-energy heavy-ion collisions are very difficult and often

More take-home messages Experimental studies of high-energy heavy-ion collisions are very difficult and often must be redone after significant improvements (resolutions, acceptances, signal to background ratios, efficiencies, etc) It is dangerous to derive a small signal by subtracting many “negligible” backgrounds Between “what you see” and “what you get”, you need some “common sense”… and the common sense changes with time and “reference frame” There is no QGP “Standard Model”… This is a data-driven field The more “explosive” is your “discovery”, the more carefully you must handle the breaks. . . 44