Bottom Quark and J/ y Production at CDF

Bottom Quark and J/yProduction at CDF Thomas J. LeCompte High Energy Physics DivisionArgonne National Laboratory For the CDF Collaboration

“Theoretical” Outline Theoretical motivation & early ideas on quarkonium production Description of the Experiment Measuring Inclusive J/y production Theoretical motivation for measuring the b cross-section Measurements at lower energy Measuring J/y’s from b quark decays Theoretical Post-dictions New Charmonium Results on the X(3872) Summary “Experimental” Outline Theoretical Ramblings A Digression The Data More theoretical ramblings A Second Digression More Data Going too fast through the last 10 or 15 slides Outline

An Introduction To Charmonium Charmonium is a bound stateof a charmed quark andantiquark. It is “almost nonrelativistic”: b ~ 0.4: Hence the hydrogen atom-likespectrum threshold 3.8 GeV 3S1 y(2S) or y’ 3P2 c2 c1 3P1 Only the most important(experimentally) statesare shown. Many morewith different quantum numbers exist. Mass 3P0 c0 J/y States can make radiative (E1) transitions to the other column. 3S1 3 GeV

Review: Quantum Numbers Spin Angular Momentum Means: Quark Spin=1 (3 = 2 x 1 + 1) Quark Orbital Ang. Mom. = 0 Total J/ySpin = 1 Orbital Angular Momentum Means: Total J/ySpin = 1 Parity is Odd Charge Conjugation is Odd Total Angular Momentum

The J/y is extremely narrow: about 87 keV: Why? Consider the possible strong decays Open charm? Nope – kinematically blocked Light quarks? Not directly – the J/y doesn’t contain any Two gluons? No Reason 1: Quantum Mechanics (Yang-Landau Theorem) – “a spin-odd particle cannot decay to two identical massless spin-1 particles” Reason 2: Violates charge conjugation symmetry Three gluons? Allowed, but suppressed In fact, electromagnetic decays compete with the strong decays! About 30% of the decays are electromagnetic/radiative The direct production rate should be tiny If J/y → gg is forbidden, so is gg → J/y Hadroproduction through c’s (followed by c → J/y + g) would be allowed This is the dominant source of J/y’s. The y(2S) rate should be really, really tiny it can’t come from c decay All y(2S) must be from the decay b → y(2S) + X Early Thinking on J/y’s Color Singlet Model OZI Rule

Theoretically: The same Yang-Landau Theorem prevents c1 production via gg interactions – but that didn’t seem to bother anybody Experimentally At fixed target energies, there is roughly the same ratio N(y(2S))/N(J/y) as at colliders This is true even at fixed target energies below b threshold! At all energies, roughly 40% (not 100%) of J/y’s come from c decays Why this is utter nonsense We should have known better. This model should have beendead on arrival; it was only the absence of alternatives that keptit going as long as it did. The field was in denial.

How Bad Was This Model? • J/y is a factor ~10 higher than predictions • That’s less bad by comparison • y(2S) is a factor ~100 higher than predictions CDF Data (20 pb-1) publishedin PRL79, 572 (1997) Even astronomers wouldcall this disagreement!

The Color Octet Model • It’s fairly clear that the CSM is missing some source of J/y’s • By the rate, it appears to be the dominant source • Consider the addition of two SU(3) (color) octets • 8+8 = 1 + 8 + 8 + 10 + 10bar + 27 • This allows 8+8 = 8: i.e. two gluons can be in a color octet state • This is analogous to the three-gluon vertex • Think of this as a two-step process • 1. The charm-anticharm pair is produced in a color octet state • 2. The octet state radiates a gluon, and becomes colorless This gets us our third gluon painlessly.Instead of ggg→ J/y, we have gg→ J/y + g This is analogous to c production: instead of a singlet c radiating a photonthere is an octet “c” radiating a gluon. The J/y Other octet states also contribute

No Free Lunch • The Color Octet Model gives us a third gluon “for free” • Because it’s soft, there is little penalty for an extra power of as • For exactly the same reason, the matrix element for the coupling between the octet c-cbar and the J/y + gluon is non-perturbative • It must be fit from experiment • All is not lost • There are only a small number of non-perturbative parameters • While they have to be fit from experiment, they have to be consistent across different measurements • There is at least one other prediction – J/y’s show a large spin-alignment at large pT Strictly speaking, the COM accommodates a largecross section – it doesn’t predict it.

Fitting COM Parameters A consistent set of COM parameters can predict reproduceboth the measured J/y and y(2S) cross-sectionsA major success of the model!

Theoretical Summary &Experimental Strategy • Color singlet prediction of 3S1 charmonium production is low by orders of magnitude • Other models can explain this, but not really predict it • Collider measurements see only the top ~6% or so (in pT) of the cross-section • NLO production of bottom quarks is also low by a factor of 2 or 3 • More details on this later in the talk • A substantial (10-20%) source of J/y’s is from b decay • Collider measurements see only the top ~10% or so (in pT) of the cross-section • Experimental plan: measure both cross-sections at ALL pT’s using the decay J/y→ mm • This will settle the issue of the b cross-sections • Experiment will be ahead of theory for the J/y and y(2S).

The CDF Detector: All you need to know Central Muon (CMU) detectors:2304 wire chambers Central Calorimeter: For this analysis, it’s used as passive steel, lead and plastic absorbers (4.7l) Open cell tracker: wire Chamber in 1.5T magneticfield (COT) Silicon vertex detector (SVX) – five layers for Precision track measurement Beams-eye view of CDF

The CDF Detector:More Than You Need To Know Silicon Vertex Detector being installed CDF rolling into the collision hall (uphill both ways)

Triggering in Words • Triggering is the key to hadron collider physics • You can’t analyze an event you didn’t trigger on (and thus record) • Collision rate (when this data was taken) is ~106 Hz • Event recording rate is ~100 Hz • Need to reject 99.99% of events • CDF Uses a 3 Level Trigger • Level 1: • Identify muon “stubs” (short tracks in the muon chambers) • Identify tracks in the transverse plane in the COT tracker with the XFT • XFT = eXtremely Fast Tracker • Level 2: • At the time this data was taken, Level 2 was in auto-accept mode for muons • Level 3: • A fast version of offline reconstruction is done • Tracks are required to have a good r-f match to the stubs • Tracks are required to have a coarse r-z match to the stubs • We don’t match east-going muons with west-going tracks • Certain kinematic cuts are made

Triggering In Pictures Two stubs in the muon chambers Two tracks in the XFTpT > 1.5 GeV Level 1 A good match between them (nominally 5 degrees) Mass between 2.6-4.0 GeVOpposite chargeGood match in r-fplaneFair match in r-z plane Level 3 (This event can’t really be a J/y – it’s shown for illustrative purposes only)

Measuring the Cross-Section • Ingredients • Number of J/y’s • Integrated Luminosity • Detector Acceptance • Detector & Trigger Efficiency • Product of several sub-efficiencies: Level-1 trigger, Level-3 trigger, tracking and muon reconstruction I will attack the denominator first

Luminosity • We used 39.7 ± 2.7 pb-1 of data in this measurement • At the time we started, this was the largest single contiguous chunk of data with common trigger conditions (February-October 2002) • Even this is broken into two pieces • 24 pb-1 taken with Df(mm) < 129o required in the trigger • Kills low pT J/y’s (oops!) • 15 pb-1 taken with this cut removed • Uncertainty is due to uncertainties at every step of the chain • Connecting our luminosity counter response to the total inelastic cross-section • Connecting the total inelastic cross-section to the total elastic cross-section (strictly speaking, the imaginary part of the forward scattering amplitude) • Connecting the total elastic cross-section to the QED Coloumb part of the elastic cross-section, which is calculable • Theoretical extrapolation between 1800 GeV (where many of the measurements have been taken) and 1960 GeV (the Run II energy) • After all this, ± 5.9% is what we end up with

Level 1 Trigger Efficiency • We also have a one-muon trigger with somewhat different requirements than the dimuon trigger • J/y events that pass this trigger have an unbiased second leg • We see how often this second leg does pass the trigger vs. pT • Note that this efficiency is for events that pass all subsequent analysis cuts

Other Efficiencies • The Level 3 and online muon reconstruction code is identical – so the efficiencies are 100% correlated • Inefficiencies vary with pT and average 1.4 ± 1.0% • Inefficiencies are due to events failing the tight (3s) track-stub matching • Failing muons either • have an early wide-angle scatter as the enter the absorber • scatter more often than typical muons • Efficiency is determined by relaxing this requirement and counting the J/y’s that have one leg fail • Offline Tracking Efficiency • Measured by embedding Monte Carlo tracks in data events and extracting them again • Efficiency is 99.6% (+0.4%, -0.9%) • Results are consistent with W → en events that come in on a trackless trigger • The Level 3 Tracking efficiency is measured like the L1 efficiency • One unbiased leg • Efficiency is 99.7 ± 0.1 ± 0.2%

Acceptance Calculation • Use a Monte Carlo with just the J/y tracks • Any confusion with the rest of the event has been taken out already in the tracking efficiency • Apply the same geometric requirements to the MC as in the data • Dead region near z=0 in the central tracker excluded • Inefficient muon wedge (HV problems with field shaping) excluded • Minor trigger error with one trigger card modeled and included

Is that bump at low pT real? high pT ~zero pT The threshold for a muon to penetrate the steel is 1.44 GeV, and the threshold to pass the triggeris 1.5 GeV – both close to ½ the J/ymass. At rest, both muons are above threshold. At high pT, both muons are above threshold. With just a little boost, though, one muon is usually below threshold and ranges out. low pT

Acceptance vs. rapidity • Our acceptance in rapidity is driven by the length of the muon chambers (± 0.6 units) • There is almost no correlation between acceptance in pT and in y. • We calculate the acceptance in 2-D bins of pT and y anyway.

Acceptance vs. J/y “polarization” • J/y’s are always produced unpolarized (<Sz> = 0) • They can, however, have “alignment” or “tensor polarization” • i.e. the density matrix is not equally populated • (<Sz2> > 0) • Gives a 1+a cos2(q) distribution • Symmetry of the J/y decay is a function of q, so alignment affects the acceptance • Affects when the softer muon ranges out • We use a = 0.15 ± 0.3 • Mix of prompt and bottom J/y’s • This corresponds to a 5-10% effect on the acceptance Run I Data

J/y Signal We have hundreds of thousands of events; we will not be statisticslimited except at the very highestpT bins. Raw pT spectrum

J/y yield in selected pT bins 12 < pT < 14 GeV 5.0 < pT < 5.5 GeV pT < 250 MeV (lowest bin) Yield is fit in each bin, corrected for acceptance and efficiency,and the cross-section bin-by-bin is calculated.There is very little feeddown from bin to bin (because the resolution is good and our bins are narrow) but we do correct for it.

Systematic Uncertainties The pT dependentterms tend to belargest at very smalltransverse momenta: The first few bins. pT dependent pT independent Combined: about 7% systematic uncertainty

The J/y Cross-Section (for |y| < 0.6)

The J/y cross-sectionin terms of pT2 Results are in excellent agreement with Run I, where there is overlap (pT > 4 GeV) (both in normalization and shape)

Turning to b production… But First, A Little History Sherman, set the Wayback Machine for 1989.

Ancient History: The Stone Age (1989) • P. Nason, S. Dawson and R.K. Ellis calculated the heavy flavor cross-section and found it to be in agreement with UA1 measurements at 630 GeV. • See Nucl. Phys. B327, 49

Understanding the x-axis:pT(min) • Ideally, one would like to measure the differential cross-section ds/dpT. • Allows comparison with theory in magnitude and shape of the cross-section. • If this is difficult, one could quote just the total cross-section. • Many experiments are insensitive to the cross-section below a pT threshold. • It makes no sense to quote the total cross-section if you have no acceptance to anything below (e.g.) 10 GeV, where the bulk of the cross-section is. • To deal with this, experiments quote the cross-section at a certain pT(min): the point where 90% of the b’s lie above. • This 90% is pure convention – we could have picked some other number • We had to pick something, so we follow the UA1 convention

Ancient History:The Bronze Age (1992) • At DPF92, CDF reported bottom quark cross-sections a factor of at least two greater than theory. • This was at a center of mass energy of 1800 GeV.

A Jump Ahead to 1997 • More recent CDF measurements show the same difficulty – the theory underpredicts the data by the same factor • This problem is not going away • Note that we measure only the high pT tail of the cross-section • Most b’s were invisible to us.

Commentary on measuring the top 10% of something Just how important could the other 90% be anyway?

Questions one might ask • Is the cross-section rising with center-of-mass energy faster than we expect? • If we take the measurements at face value, that’s what we would conclude • Not a completely crazy idea • the NLO contributions are larger at 1800 GeV than 630 GeV • The scale dependence of the calculation is worse at NLO than LO • This is due to a numerical accident, but was not widely known at this time • Large NNLO contributions might produce additional growth with center-of-mass energy • Did (at least) one experiment get the measurement wrong? • Is something wrong with our theoretical models? • Extra b sources? (ANL group) • Fragmentation? (P. Nason et al.) • Is there anything we can do to put the experimental result on a more solid footing?

The Enlightenment (1995-6) • In the winter of 1995-6, we ran for 9 days at 630 GeV to address this question. • We estimated 50 to 100 b’s at the lower energy, depending on whether this factor of 2 was real or not. • Not many, but enough to measure a factor of 2 • At the same time, we could collect jets and photons and do other QCD measurements • Several Ph.D. theses have resulted from these measurements, and they were important in untangling the high ET jet excess • Eleven papers were published by CDF and D0 based on this data • This run was proposed and largely executed by ~6 people

The Dark Ages: 1995-2000 &The Renaissance: 2001-2002 • Work beyond the preliminary stages stopped – the CDF upgrade expanded to consume all available time • 4 Lehman reviews • $3.6 million dollars • 15 change requests • innumerable monthly reports • Yellowing scintillator • Not yellow like Coors beer • Yellow like a lemon • shady vendors • squabbling collaborators… • Only when the upgrade was behind us did this start moving again – see Phys.Rev.D66:032002,2002 And this was just the muon upgrade!

The Measurement: Some Key Ideas • We measure a ratio of cross-sections because it is both theoretically better determined and experimentally more certain • Theory uncertainty is 10-15% rather than a factor of 2 or more • This measurement is statistically limited by the number of bottom events collected at 630 GeV • Complications to improve understanding of other aspects like acceptance or efficiency will make a minimal impact on the final answer • We took pains to make the 1800 GeV sample as similar to the 630 GeV sample as possible. • We rejected larger samples with more differences – for example, we could have used a sample that used an earlier version of our silicon detector, but we didn’t. • Every event is taken within 3 weeks of the 630 GeV run • Much of this sample is taken from the period at 1800 GeV where we tested the 630 GeV trigger table

Luminosity and Datasets • Run I CDF used a three tier trigger. For this analysis, we required • Level 1: a 6 GeV muon “stub” in the central muon chambers, plus at least 2 hits in the corresponding outer chamber • Level 2: that stub matched to a 4.7 GeV r-f track • Level 3: a 4.5 GeV muon with good matching to both the inner and outer muon chambers • Offline, we required • A 5 GeV muon with a good match between the track and the muon stub • These are the same requirements for both 630 and 1800 GeV • Integrated luminosity: • At 1800 GeV: 1932 nb-1 • At 630 GeV: 582 nb-1

Start with a beautiful muon I will spare you the details Find all the tracks with pT above 1 GeV and m(mh) < 5.3 GeV in a cone of DR < 1 around the muon Select the highest pT track Find the vertex of that track and the muon Perform quality cuts Again, I’ll spare you the details Count the excess of events with the vertex forward of the interaction point vs. behind the interaction point Finding Beautiful Hadrons hadron m muon The Goldilocks Principle: Inclusive semileptonic decays are too impure. Exclusive decays are too rare. These are just right. Lxy Interaction Point

3083 events ahead of the primary vertex (by at least 250 mm) 1527 events behind Yield is 1556 ± 68 bottom events Lifetime (as a check) is 1.4 ± 0.1ps Counting b’s at 1800 GeV

383 events ahead of the primary vertex (by at least 250 mm) 200 events behind Yield is 183 ± 24 You don’t get many b’s in a short, low energy run Lifetime is 1.4 ± 0.3 ps Counting b’s at 630 GeV

Cross-Section Ratio • We can put it all together to find the cross-section ratio • Comparison with NDE predictions and MRS-A’ parton densities is good • Other PDF’s (MRSA, CTEQ 6M) give essentially the same prediction

Comparison with UA1 • We take the CDF measured b cross-section at 1800 GeV, multiply it by the derived ratio, and place the cross-section obtained on the UA1 plot. • It shows • We are a factor of ~2 higher than NLO QCD • How could it be otherwise? • We have smaller error bars than UA1 • This is the best single measurement of the b cross section at these energies!

Summary of 630 GeV Run • CDF is marginally consistent with UA1 • Reject UA1 at 90% confidence level • Fail to reject UA1 at 95% confidence level • The CDF central value is above theoretical predictions by a similar factor at 630 GeV as at 1800 GeV • There is no indication as to why this is • But we can exclude the cross section growing with center of mass energy • Precision measurements are possible in heavy flavor production experiments • Uncertainties of ~15%, not factors of 2 • It’s a heckuva lot of fun to propose and run your own small experiment • And 11 papers out of 9 days of running is not too shabby…

Back to 1960 GeV:The b cross-section using J/ys • Basic strategy: • We know the J/y cross-section • We know the branching fraction of b’s to J/y’s (about 1.1%) • If we can measure the fraction of J/y’s from b’s, we’re one multiplication and one division away from the b quark cross-section • Lifetime is the key • B hadrons live ~1.5 ps • We can use the SVX (silicon vertex detector) to identify J/y’s that were not produced at the primary vertex – these must be from b decay. • Complications • The most probable decay time is zero: some b’s are identified as non-b’s • Our measurement is not perfect: some non-b’s are identified as b’s. • Slowly moving b hadrons don’t get very far before they decay • Separation power is poor at low pT

Two acceptance complications • For us to separate prompt and non-prompt b’s accurately, we need to impose tight silicon requirements • No more than one hit missed • At least three hits • Avoid “bad regions” – e.g. crossing silicon barrels • Since we have some dead silicon ladders, these requirements may sculpt the acceptance (only about 1 in 3 J/y’s have both muons pass these criteria) • Events where we can measure the b fraction may not be representative of unbiased J/y’s. • We checked this, and the acceptance ratio (good silicon/total) is flat in pT(J/y) • The spin alignment parameter a is different for b’s and inclusive J/y’s • This means the acceptance is different • We have to (and do) correct for this: it’s a ~10% effect • For prompt J/y’s we use our Run I measurement • For b’s, we take the (better & more recent) BaBar measurement and boost into our frame • This is not entirely trivial, since BaBar measures this in the U(4S) frame • a(BaBar) = -0.09 ± 0.10

p(y) m+ m+ p(B) q Rxy Lxy Choice of Separation Variable • Variables of Interest • Rxy • The transverse flight distance • Has the best separation power • Lxy • The transverse flight distance dotted into the unit y momentum vector • Differs from Rxy by cos(q) • A signed quantity • Pseudo-ct • Lxy boosted to B rest frame based on average boost derived from MC • Differs from Lxy by a known multiplicative constant • ct • The true B lifetime • Differs from Rxy by an unknownmultiplicative constant We use Lxy/pT: we trade statisticalseparation power for better control over systematics. The “/pT”corrects for the Lorentz boost

A word on b decay kinematics • Above 2 GeV, <pT(b)> is proportional to <pT (y)> • Below 1.5 GeV, <pT(b)> is more or less constant • This is because <pT(y)> is driven largely by the b decay kinematics, not by the b production dynamics • The <Rxy> distribution looks like this as well (it has to) • The <Lxy> distribution looks qualitatively like this • Once we get to J/y’s of pT < 1.7 GeV or so, we are probing b’s down to pT’s of 0.

Fitting the B fraction • Prompt component • Resolution function is determined from the zero-lifetime component • Double-Gaussian with some small tails at large negative lifetime • B component • Exponential convolved with the resolution function determined from the prompt component • Sidebands • We assume the background under the J/y mass peak is modeled by the weighted average of the sidebands • Note that there are B’s (double semileptonic decays) in the sidebands

Bottom Quark and J/ y Production at CDF