FIRST OBSERVATION OF THE ηb MESON AND STUDY OF THE
DECAYS Υ (3S) γηb AND Υ (2S) γηb → →
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF PHYSICS
AND THE COMMITTEE ON GRADUATE STUDIES
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Christopher West May 2010
© 2010 by Christopher Alan West. All Rights Reserved. Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/nv861zm8623
ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Rafe Schindler, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Patricia Burchat
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Jonathan Dorfan
I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Michael Peskin
Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education
This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.
iii Abstract
This dissertation presents the results of an analysis of data samples consisting of 122 million Υ (3S) decays and 90 million Υ (2S) decays collected with the BABAR detector operating at the PEP-II asymmetric-energy storage rings at SLAC National Accelera- tor Laboratory. The η meson is observed for the first time in the decay Υ (3S) γη b → b and its existence is confirmed in the decay Υ (2S) γη . The η mass extracted → b b from the Υ (3S) γη sample is 9388.9+3.1(stat) 2.7(syst) MeV/c2, correspond- → b −2.3 ing to a hyperfine mass splitting m m = 71.4+2.3(stat) 2.7(syst) MeV/c2. Υ (1S) − ηb −3.1 The branching fraction for the decay Υ (3S) γη is determined to be [4.8 → b 0.5(stat) 0.6(syst)] 10−4. An analysis of the decay Υ (2S) γη confirms the × → b
observation of the ηb meson and provides an additional measurement of the ηb mass: 9394.2+4.8(stat) 2.0(syst) MeV/c2. The branching fraction for the decay Υ (2S) γη −4.9 → b is determined to be [3.9 1.1(stat)+1.1(syst)] 10−4. As the measurements in the −0.9 × two samples are consistent, the masses are averaged to provide a combined value of m = 9390.9 3.2 MeV/c2. After canceling common systematic errors, the following ηb branching fraction ratio is determined: [Υ (2S) γη (1S)]/ [Υ (3S) γη (1S)] = B → b B → b 0.82 0.24(stat)+0.20(syst). −0.19
iv Acknowledgments
A Ph.D. dissertation is never completed without the help of many other people, and mine is no exception. First of all, I would like to thank my adviser, Rafe Schindler, for his guidance throughout the years and giving me the freedom to discover my
interests in physics at BABAR. A great deal of my knowledge of physics analysis comes directly from those with whom I have worked: Philippe Grenier, Peter Kim, Peter Lewis, Silke Nelson and Veronique Ziegler. In particular, I would like to thank
Peter Kim for spurring my interest in searches for undiscovered states, and the ηb in particular. My analysis of radiative decays benefits from the knowledge I gained in electromagnetic calorimetry from Martin Kocian and Bill Wisniewski. The support of my Group E colleagues Walt Innes, Peter Lewis, Selina Li, Martin Perl and Stephen Sun made the task of completing my dissertation a pleasant experience. Finally, I would like to thank my parents, Bill and Claudia West, for always believing in me
and encouraging me to pursue my dreams.
v Contents
Abstract iv
Acknowledgments v
1 Introduction 1
1.1 The Standard Model ...... 2
1.2 Quantum Chromodynamics ...... 3
1.2.1 Historical Overview ...... 3
1.2.2 Asymptotic Freedom ...... 4
1.3 Theoretical Approaches to QCD ...... 5
1.3.1 Quark Potential Models ...... 5
1.3.2 Lattice QCD ...... 6
1.3.3 Non-relativistic quantum chromodynamics ...... 10
1.3.4 Potential non-relativistic quantum chromodynamics ...... 12
1.4 Bottomonium ...... 13
1.5 Electromagnetic transitions ...... 13
1.6 The ground state of bottomonium, the ηb meson ...... 15
1.7 Conclusion ...... 22
vi 2 The BABAR Detector 24
2.1 Physics Motivation ...... 24
2.1.1 CP violation ...... 24
2.1.2 Bottomonium Physics ...... 26
2.2 PEP-II ...... 26
2.3 The BABAR Detector ...... 26
2.4 Silicon Vertex Tracker ...... 27
2.5 Drift Chamber ...... 30
2.6 Detector of Internally Reflected Light ...... 34
2.7 Electromagnetic Calorimeter ...... 36
2.7.1 Design ...... 37
2.7.2 Calibration and Performance ...... 38
2.8 Superconducting Coil ...... 41
2.9 Instrumented Flux Return ...... 42
2.10 Trigger ...... 43
2.10.1 Level 1 Trigger ...... 44
2.10.2 Level 3 Trigger ...... 45
2.11 Datasets ...... 45
3 Analysis Overview 47
3.1 Introduction ...... 47
3.2 Backgrounds ...... 48
3.2.1 Non-peaking ...... 48
3.2.2 Υ (nS) γχ (mS), χ γΥ (1S) ...... 49 → b b → 3.2.3 e+e− γΥ (1S) ...... 49 → vii 3.3 Selection Criteria ...... 50
4 Study of the Decay Υ (3S) γη 52 → b 4.1 Selection Criteria ...... 52
4.1.1 Data set ...... 52
4.1.2 Event selection ...... 53
4.1.3 π0 veto optimization ...... 55
4.1.4 Investigation of a possible η veto ...... 62
4.1.5 N-1 cut plots ...... 64
4.2 Background to the Eγ spectrum ...... 66
4.2.1 Introduction ...... 66
4.2.2 Non-peaking Background ...... 67
4.2.3 Peaking Background from χbJ (2P ) → γΥ (1S) ...... 69
+ − 4.2.4 Peaking Background from e e → γISRΥ (1S) ...... 73
4.3 Fitting procedure ...... 81
4.3.1 Fit and Unblinding strategies ...... 81
4.3.2 Minimization Details ...... 82
4.4 Toy Studies ...... 84
4.4.1 Conclusions ...... 86
4.5 Fit on the 2.5 fb−1 Optimization Sample ...... 93
4.6 Fit to the inclusive photon spectrum ...... 93
4.7 Systematic Uncertainties ...... 96
5 Study of the Decay Υ (2S) γη 98 → b 5.1 Selection ...... 98
viii 5.2 Background Modeling ...... 102
5.2.1 Introduction ...... 102
5.2.2 Non-peaking Background ...... 102
5.2.3 e+e− γ Υ (1S) ...... 103 → ISR 5.2.4 Υ (2S) γχ (1P ), χ (1P ) γΥ (1S) ...... 114 → bJ bJ → 5.2.5 Υ (2S) Υ (1S)(η, π0) ...... 117 → 5.2.6 Υ (2S) Υ (1S)π0π0 ...... 119 → 5.3 Control Sample Studies ...... 122
5.3.1 Selection ...... 122
5.3.2 Monte Carlo ...... 123
5.4 Fit Procedure for Υ (2S) γη Sample ...... 129 → b 5.4.1 Introduction ...... 129
5.4.2 Fixed and floating parameters ...... 130
5.4.3 Fit to test sample, ISR yield fixed ...... 131
5.4.4 Fit to test sample, ISR yield floating ...... 134
5.4.5 Fit to full sample with signal region blinded, ISR yield fixed . 139
5.4.6 Fit to full sample with signal region blinded, ISR yield floating 139
5.5 Toy studies ...... 146
5.5.1 Introduction ...... 146
5.5.2 Toys with ISR yield fixed to incorrect yield ...... 147
5.5.3 Conclusion ...... 148
5.6 Fit results ...... 149
5.7 Systematic Errors ...... 155
5.7.1 Sources of Systematic Errors ...... 155
ix 5.7.2 Significance of Signal (Including Systematic Errors) ...... 156 5.7.3 Additional Fit Variations ...... 157
5.7.4 Branching Fraction Uncertainties ...... 160 5.8 Combination with Υ (3S) result ...... 164 5.8.1 Ratio of branching fractions ...... 164 5.8.2 New Υ (3S) γη branching fraction ...... 167 → b
5.8.3 Mass of the ηb ...... 168
6 Summary and Outlook 169
Appendices
A Use of EMC Timing to Improve π0 Veto 171
B Spurious Feature at 680 MeV 176 B.1 Description of problem ...... 176 B.2 Other investigations of spike ...... 177
Bibliography 183
x List of Tables
1.1 Predictions for the hyperfine splitting m m from lattice QCD Υ (1S) − ηb and perturbative QCD calculations...... 17
1.2 Predictions for the two-photon partial decay width, taken from [41],
and the total width determined by scaling the two-photon decay width using Eq. 1.12...... 20
1.3 Limits (95% confidence level) on ηb two photon partial width times branching fraction from LEP...... 23
4.1 On-resonance datasets used in the selection optimization ...... 53
4.2 Selection efficiencies () for truth-matched signal MC and on-peak data
in the energy range 0.85 < Eγ < 0.95, in percent. The reconstruction efficiency on data is normalized to 100%...... 56
4.3 Comparison of single cut efficiencies and S/√B from fitted χb yields and truth-matched signal MC. The background contribution is found
by integrating the background function from E 2σ to E + 2σ. 59 g,χb1 − g,χb2 xi 0 4.4 Signal to background study for the π veto. nχb and ∆nχb are the
fitted χb signal yield and its error. B is the background determined by integrating the background function from E 2σ to E + 2σ. g,χb1 − g,χb2
S/√B of the χb peaks is derived from the fitted χb yield and the
background underneath the peaks. The background in the ηb search region, is the integral of the background distribution in the search
region, defined by 0.85 < pγ < 0.95...... 60
4.5 Signal to background study for π0 veto with additional 3σ timing cut.
The results of the timing from the simulation are not quoted, as the simulation matches the data poorly...... 61
4.6 Fitted background parameters from a fit of the full data sample in the
side region 0.5 < Eγ < 0.6 GeV and 0.960 < Eγ < 1.2 GeV...... 69
4.7 Summary of χbJ (2P ) transitions. In the last column we give the num- ber of expected events, corrected for efficiency, for 30 fb−1 of on reso- nance data...... 69
4.8 Fitted parameters for the χbJ (2P ) peaks in the blinded full dataset spectrum. A, N and σ are the transition point, tail parameter, and Gaussian width, respectively, of the Crystal Ball function. The offset is defined by E E ...... 71 γ,expected − γ,measured
4.9 Production cross section for e+e− γ Υ (1S) at √s = 10.3252 GeV → ISR (σ ), production cross section for e+e− γ Υ (1S) at √s = Υ (3S) → ISR
10.55 GeV (σΥ (4S)), and their ratio for various orders in perturbation theory. The assumed di-electron width of the Υ (1S) is 1.340 MeV. . . 75
xii 4.10 Number of e+e− γ Υ (1S) events from the Υ (3S) off-resonance → ISR and Υ (4S) off-resonance samples, and extrapolation to the Υ (3S) on-
resonance sample (25.598 fb−1). The errors are statistical only. . . . . 76
4.11 Fitted peak values (µ) and width (σ) of ηb peak position value distri-
bution in GeV for various combinations of ηb yield (in thousands) and photon peak positions...... 85
4.12 Mean and width of signal significance distributions for various combi-
nations of ηb yield (in thousands) and photon peak positions. . . . . 86
4.13 Signal yields and width (σ) of signal yield distribution for various com-
binations of ηb yield (in thousands) and photon peak positions. . . . . 87
4.14 Mean and width of the fitted ηb peak positions for different yields of ISR events ...... 88
4.15 Mean and width of the fitted ηb signal significance for different yields of ISR events ...... 88
4.16 Mean and width of the fitted ηb signal yield for different yields of ISR events ...... 89
4.17 Mean and width of the fitted ηb peak position in GeV for different
generated and fitted ηb widths. Generated peak position is 910 MeV and generated yield is 20k events...... 90
4.18 Mean and width of the fitted ηb signal significance in GeV for different
generated and fitted ηb widths. Generated peak position is 910 MeV and generated yield is 20k events...... 91
xiii 4.19 Mean and width of the fitted ηb signal yield in GeV for different gen-
erated and fitted ηb widths. Generated peak position is 910 MeV and generated yield is 20k events...... 92
4.20 Results of the fit to the Eγ distribution assuming the nominal ηb width
of Γ(ηb) = 10 MeV...... 93
4.21 Results of the fit to the Eγ distribution with different assumed values
of Γ(ηb)...... 94
4.22 Systematic uncertainties for the measurement of the ηb yield and peak position (mass)...... 97
4.23 Systematic uncertainties for the measurement of the ηb branching frac- tion...... 97
5.1 Improvement of S/√B as the track and lateral moment selections are imposed. The overall normalization is arbitrary...... 99
5.2 Cut efficiencies for truth-matched signal MC determined by counting
events in the energy range 0.3 < Eγ < 0.7 GeV...... 100
5.3 Systematic errors on extrapolation of off-Υ (4S) ISR rate to on-Υ (2S) data...... 106
5.4 Selection efficiencies for truth-matched e+e− γΥ (1S) signal MC de- → termined by counting events, at Υ (4S) off-peak and Υ (2S) on-resonance
energies. The photons are counted in the energy ranges 0.4 < Eγ <
0.6 GeV and 0.85 < Eγ < 1.10 GeV for the Υ (2S) on-peak and Υ (4S) off-resonance energies, respectively...... 108
xiv 5.5 Number of e+e− γ Υ (1S) events from the Υ (2S) off-resonance → ISR and Υ (4S) off-resonance samples, and extrapolation to the Υ (2S) on-
resonance sample (13.4 fb−1), excluding the test sample. The errors are statistical only. We calculate that at the peak of the Υ (2S) resonance the cross section for radiative return to the Υ (1S) is 41.3 pb. The efficiencies for the Υ (2S) and Υ (3S) off-peak are determined from the
corresponding on-peak MC dataset...... 113
5.6 Branching fractions for the decays Υ (2S) γχ (1P ), χ (1P ) → bJ bJ → γΥ (1S), and the CM energies of the photon in the first and second decay. The values are taken from the PDG...... 114
5.7 Theoretical branching fractions for the decays Υ (2S) γχ (1P ), χ (1P ) → bJ bJ → γΥ (1S), and the CM energies of the photon in the first and second de- cay from [49]. In deriving the branching fractions from the radiative width, we use a full width of the Υ (2S) Γ = 31.98 2.63 keV. . . . . 115
5.8 Width of χcJ , J = 0, 1, 2 states from the PDG [79]. The widths of the
χbJ are expected to be smaller than those of the χcJ . This provides
the justification for neglecting the χb width in the fit...... 115
5.9 Doppler broadening due to momentum of χbJ relative to the CM frame. The value listed is the half-width of the Doppler broadening PDF. . . 117
xv 5.10 Result of simultaneous fit to χbJ , J = 0, 1, 2 MC. The peak position, Doppler width and signal yield are labeled by their total angular mo-
mentum, J. The remaining parameters are shared between the three peaks. They are A, the transition point between the Gaussian and power tail components of the Crystal Ball function, in units of σ from the mean of the peak, and N, the power law parameter. Note that
there are 96,000 events in the J = 1 MC sample compared to 145,000 events in each of the J = 0, 2 data samples which must be taken into account in computing the efficiency...... 118
5.11 Cut efficiencies determined by counting events in χbJ (1P ) MC. . . . . 119
5.12 Parameters resulting from fits to χbJ MC...... 124
5.13 γsoft spectrum fitted parameters, final fit...... 128
5.14 Parameters from fit to test sample with the ISR yield fixed. The sec-
tions of the table are yields, the ηb peak position, background param-
eters, and χb lineshape parameters. The ISR yield is fixed in the fit. . 134
5.15 Parameters from fit to optimization sample with the ISR yield uncon-
strained. The sections of the table are yields, the ηb peak position,
background parameters, and χb lineshape parameters...... 138
5.16 Parameters from fit to full sample with signal yield blinded. The sec-
tions of the table are yields, the ηb peak position, background param-
eters, and χb lineshape parameters. The ISR yield is fixed to 17324, a value determined from an earlier study of the ISR yield...... 141
xvi 5.17 Fit parameters from fit to full sample with signal yield blinded and
floating ISR yield. The sections of the table are yields, the ηb peak
position, background parameters, and χb lineshape parameters. . . . . 145
5.18 Signal yields, yield errors and pulls from toy studies with ηb width fixed at 5 MeV. We show both the mean (µ) and width (σ) from a Gaussian fit to the pull distribution...... 147
5.19 Signal peaks, peak errors and pulls from toy studies with ηb width fixed at 5 MeV. We show both the mean (µ) and width (σ) from a Gaussian
fit to the pull distributions...... 147
5.20 Signal yields, yield errors and pulls from toy studies with ηb width fixed at 10 MeV. We show both the mean (µ) and width (σ) from a Gaussian fit to the pull distribution...... 147
5.21 Signal peaks, peak errors and pulls from toy studies with ηb width fixed at 10 MeV. We show both the mean (µ) and width (σ) from a Gaussian
fit to the pull distributions...... 148
5.22 Fit parameters from fit to full sample with floating ISR yield. The
sections of the table are yields, the ηb peak position, background pa-
rameters, and χb lineshape parameters...... 150
2 5.23 Signal yields and χ values for fits with alternate ηb widths...... 152
5.24 Systematic uncertainties for the measurement of the ηb yield and peak position (mass). Errors with no sign given are taken to be symmetric. 157
5.25 Yield and signal significance (statistical only) with several fit variations.159
5.26 Comparison of selection efficiencies in χbJ (2P ) data versus χb0(2P ) MC.160
5.27 Systematic uncertainties on the selection efficiency, in percent. . . . . 163
xvii 5.28 Systematic uncertainties for the measurement of the ηb branching frac- tion, in percent...... 163
5.29 Yield of ηb signal in Υ (2S) and Υ (3S) analyses, the ratio of the two yields, and the deviation of the ratio from that of the 10 MeV fit, as a
function of the ηb width assumed in the fit. Υ (2S) and Υ (3S) are used as abbreviations for the Υ (2S) and Υ (3S) yields, respectively. . . . . 164
5.30 Systematic uncertainties (in %) on the selection efficiency in the Υ (3S) → γη and Υ (2S) γη analyses, in percent. Systematics which are as- b → b sumed to cancel in the branching fraction ratio are in italics...... 166 5.31 Systematic uncertainties on the branching fraction ratio, in percent. . 166
5.32 Systematic uncertainties on the new Υ (3S) γη branching fraction, → b in percent...... 167
xviii List of Figures
1.1 Cornell model potential for σ = 0.18 GeV2...... 7
1.2 Spectrum of b¯b levels from a recent lattice prediction [24]. Closed
and open symbols are from coarse (lattice spacing a 0.12 fm) and ≈ fine (a 0.09 fm) lattices, respectively. Squares and triangles de- ≈ note unquenched and quenched results, respectively. Lines represent experiment...... 8
1.3 Spectrum of b¯b levels and possible transitions between levels. Figure from [36]...... 14
1.4 Spectrum of b¯b levels from a recent lattice prediction [24]. Closed and open symbols are from coarse (lattice spacing a 0.12 fm) and ≈ fine (a 0.09 fm) lattices, respectively. Squares and triangles de- ≈ note unquenched and quenched results, respectively. Lines represent experiment...... 18
1.5 Hyperfine splitting M M as a function of renormalization scale Υ (1S) − ηb µ in leading order (dotted line), next to leading order (dashed line), leading log (dot-dashed line), and next-to-leading log (solid line) ap- proximations. For the next-to-leading log result, the yellow band cor-
responds to a variation α (M ) = 0.118 0.003. Figure from [19]. . . 19 s Z xix 1.6 Theoretical predictions for Υ (3S) γη branching fraction and CLEO → b upper limit...... 22
2.1 BABAR detector longitudinal section...... 28
2.2 Silicon vertex tracker longitudinal section...... 29
2.3 Drift chamber longitudinal section. Measurements are in mm. . . . . 31
2.4 Drift chamber wire placement for first four superlayers. Stereo angles
are in mrad...... 32
2.5 Single cell resolution as a function of distance from sense wire. . . . . 33
2.6 Specific energy loss versus momentum. Lines show predictions for var- ious mass hypotheses from the Bethe-Bloch equation...... 34
2.7 Diagram of DIRC. Distances are not to scale...... 35
2.8 Cherenkov angle and timing resolution of the DIRC...... 36
2.9 Invariant mass plot of candidate D0 Kπ events before and after → applying loose kaon ID...... 37
2.10 Longitudinal section of the EMC. Lateral dimensions are shown in mm and angles in degrees...... 38
2.11 Fractional energy resolution σE/E versus energy as derived from vari- ous sources, compared to Monte Carlo expectations...... 40
2.12 EMC angular resolution as a function of photon energy...... 41
2.13 Geometry of barrel and endcap regions of the IFR...... 42
2.14 Pion rejection versus muon efficiency for track momentum 0.5 < p <
2.0 GeV/c (left) and 2 < p < 4 GeV/c (right)...... 43
xx 4.1 Photon CM energies for truth-matched signal photons. Photons inside the calorimeter fiducial region (top left), photons inside the barrel but
outside the fiducial region (top right). The top plots shown using the
same scale (bottom left). Photon energies versus cos(θγ,LAB) (bottom right). The blue (red) points show photons accepted (rejected) by the fiducial requirement. The gradual worsening of the resolution as one
moves away from θLAB = 0 is due to the increased photon path length through the DIRC...... 55
4.2 Cluster lateral moment (left), number of charged tracks in event (right), cosine of photon momentum direction with the thrust axis of the rest
of the event (bottom), for truth-matched signal (red) and data (blue). The distributions are shown prior to the application for any selection
criteria except a requirement of 0.85 < Eγ < 0.95. The arrows show the values of the selection criteria...... 57
4.3 S/√B derived from truth-matched signal MC and on-peak data. The
on-peak data is evaluated in a region 0.85 < Eγ < 0.95...... 58
0 4.4 S/√B of χb peaks versus π second photon energy cut. With (without) timing cut shown in blue (red). The overall scale is arbitrary. Note the suppressed zero...... 61
4.5 Efficiency of η veto on data and ηb signal MC, and change in S/√B obtained by applying the veto. No π0 veto is applied. Note the sup-
pressed zero...... 62
xxi 4.6 Efficiency of η veto on data and Monte Carlo, and change in S/√B obtained by applying the veto. A π0 veto is applied. Note that the
vertical scale is different from that of Fig. 4.5...... 63
4.7 Normalized distribution of selection variables when all other cuts are applied in signal MC (blue) and data (red). The variables are, from
left to right, the number of tracks, the cluster lateral moment, cos θγ , cos θ and m ...... 65 | T | γγ 4.8 Full dataset blinded spectrum, used to determine the continuum back- ground parametrization. This histogram shows the fit while the data
is shown by the markers...... 68
4.9 Fit to the χbJ (2P ) peaks in the blinded full dataset subtracted spec- trum. The medium figure is a zoom and the bottom shows the resid- uals...... 72
4.10 ISR photon c.m. energy distribution in e+e− γ Υ (1S) signal MC → ISR (data points). The top figure shows the truth-associated spectrum, the bottom figure the total spectrum. The superimposed signal lineshape corresponds to the Crystal Ball function (red). The background (green
line) is described using a exponential function. The total fit function is represented by the blue curve...... 77
4.11 The photon c.m. energy distribution in Υ (3S) off-resonance data (data points). The superimposed ISR-signal lineshape corresponds to the Crystal Ball function (red). The background (green line) is described by a function of the form given in Eq. 4.2. The total fit function is
represented by the blue curve...... 78
xxii 4.12 Shown in the top two plots is the photon c.m. energy distribution in e+e− γ Υ (1S) signal MC, generated at the Υ (4S) off-resonance → ISR energy. The photon c.m. energy distribution in the Υ (4S) off-resonance data sample is shown on the bottom. The ISR-signal lineshapes are shown in red(top) and black(bottom). The background (dashed-blue on top, green line on bottom) is described using an exponential func-
tion (top) or a function of the form given in Eq. 4.2 (bottom). The total fit functions are represented by the blue curves...... 79
4.13 Inclusive photon spectrum in the below-Υ (4S) data, after background subtraction. The fitted curve shown is the Crystal Ball Function which describes the data points very well...... 80
4.14 The inclusive photon c.m. energy distribution in the optimization sam-
ple (2.5 fb−1) of Υ (3S) on-resonance data. The top plot show the dis- tribution together with the fit results and the bottom plot shows the non-peaking background subtracted plot...... 94
4.15 (a) Inclusive photon spectrum in the region 0.50 < Eγ < 1.1 GeV. (b) Background subtracted photon spectrum in the signal region, showing
χbJ (2P ) peaks (red), ISR Υ (1S) (green), signal (blue) and the sum of the contributions (purple). (c) Signal peak after all backgrounds are subtracted...... 95
5.1 Number of truth-matched signal photons (top left) and background photons (top right) passing all cuts as the cuts on cos θ and E are | T | γ2 varied. S/√B (bottom) is computed using these two quantities. The
overall normalization is arbitrary...... 101
xxiii 5.2 Spectrum plotted with logarithmic y-axis...... 103
5.3 Fit of photon energy spectrum in ISR Υ (1S) MC at Υ (2S) energy using truth-matched (top) and all signal candidate photons (bottom). The blue curve is the Crystal Ball function used as the signal PDF and the red line is the polynomial background...... 104
5.4 Fit of MC photon energy spectrum in ISR Υ (1S) signal region in MC generated at Υ (4S) off-resonance energy. The signal PDF is a Crystal Ball function and the background PDF is an exponential...... 106
5.5 Fit of photon energy spectrum for ISR Υ (1S) production in off-resonance Υ (4S) data before (top) and after (bottom) subtracting the smooth background. The green (blue) curve in the top plot is the background (total) fit function. The normalized residuals from the fit are shown in the middle plot...... 109
5.6 Fit of photon energy spectrum of ISR Υ (1S) production in off-resonance Υ (2S) data before (top) and after (bottom) subtracting the smooth background. The green (blue) curve in the top plot is the background
(total) fit function. The normalized residuals from the fit are shown in the middle plot...... 110
5.7 Fit of MC photon energy spectrum in ISR Υ (1S) signal region in Υ (3S) off-peak MC. The signal PDF is a Crystal Ball function and the back-
ground PDF is an exponential...... 111
xxiv 5.8 Fit of photon energy spectrum for ISR Υ (1S) production in off-resonance Υ (3S) data before (top) and after (bottom) subtracting the smooth
background. The green (blue) curve in the top plot is the background (total) fit function. The normalized residuals from the fit are shown in the middle plot...... 112
5.9 Ratio of histograms of photon energy spectrum of χb2 decays before/after
a 25 cm bump distance cut (black). The red curve is the χb2 photon spectrum. Plot courtesy of Steve Sekula...... 116
5.10 Fit of zero-width χb0 (top), χb1 (middle), and χb2 (bottom) MC using truth-matched photons...... 120
5.11 Spectrum of truth-matched photons from the decays Υ (2S) ηΥ (1S)(top) → and Υ (2S) Υ (1S)π0π0 (bottom). Note that the x-axis is different →
from that of the final fit, which is over the region 270 < Eγ < 800 MeV. 121
5.12 Photon spectrum before cuts for double radiative decay of Υ (2S) can- didates. γ window (red): 60 200 MeV, γ window (blue): soft − hard 380 470 MeV...... 123 −
5.13 Monte Carlo χb0 peak with fit superimposed. The curves displayed on the plot are the exponential background (short dotted light blue line), the signal Gaussian (dot-dash green line), the signal Crystal Ball function (red dashed line) and their sum (solid blue line). χ2 per degree of freedom = 193/91 ...... 124
5.14 Monte Carlo χb1 peak with fit and normalized residual plot. The curves are as in Fig. 5.13. χ2 per degree of freedom = 94/81 ...... 125
xxv 5.15 Monte Carlo χb2 peak with fit and normalized residuals plot. The curves are as in Fig. 5.13. χ2 per degree of freedom = 42/61 . . . . . 126
5.16 Fit to soft transitions in data with normalized residuals plot. The exponential background function is shown as a short dotted light blue line, each Gaussian is a dot-dash green line, the red dashed lines are the Crystal Ball functions and the solid blue line is the total fit. χ2/ndof =
123/131...... 128
5.17 Fit to the test sample (top) and residuals from the fit (bottom). The ISR yield is fixed in the fit...... 132
5.18 Fit to the test sample after subtracting the non-peaking background (top) and the corresponding zoomed version (bottom). The χ2/ndof is equal to 93.8/94 = 0.98, corresponding to a fit probability of 49%.
The histograms are χb (cyan), ISR (red), and ηb (blue). The ISR yield is fixed in the fit...... 133
5.19 Fit to the test sample (top) and normalized residuals from the fit (bot- tom). The ISR yield is floating in the fit...... 135
5.20 Fit to the test sample after subtracting the non-peaking background
(top) and the corresponding zoomed version (bottom). The χ2/ndof is equal to 93.7/94 = 0.98, corresponding to a fit probability of 49%.
The histograms are χb (cyan), ISR (red), and ηb (blue)...... 137
5.21 Fit to the full data sample (top) and residuals from the fit (bottom).
The ISR yield is fixed in the fit...... 140
xxvi 5.22 Fit to full data sample after background subtraction (top) and the corresponding zoomed plot (bottom). The cyan curve is the sum of
the χb peaks, and the red curve is the ISR peak. The histograms are
χb (cyan) and ISR (red). The ISR yield is fixed in the fit...... 142
5.23 Fit to the full data sample (top) and residuals from the fit (bottom). 143
5.24 Fit to full data sample after background subtraction (top) and the
corresponding zoomed plot (bottom). The ηb signal region is excluded
in the residual plot. The cyan curve is the sum of the χb peaks, and
the red curve is the ISR PDF. The histograms are χb (cyan) and ISR (red). The ISR yield is floating in the fit...... 144
5.25 Fit to the full data sample (top) and residuals from the fit (bottom). 151
5.26 Fit to full data sample after smooth background subtraction (top), the corresponding zoomed plot (middle) and after all backgrounds are
subtracted (bottom). The cyan curve is the sum of the χb peaks, and
the red curve is the ISR PDF. The histograms are χb (cyan), ISR (red),
ηb (blue), and their sum (black). The ISR yield is floating in the fit. . 153
5.27 Same as fig. 5.26 but with the ηb PDF removed...... 154
5.28 Fit to test sample before π0 veto. The cyan curve is the sum of the
χb peaks, and the red curve is the ISR PDF. The histograms are χb
(cyan), ISR (red), and ηb (blue). The ISR yield is floating in the fit. . 162
A.1 Deviation from average timing before cluster timing correction. The 48 ns shift from the long shaping time pre-amplifiers has already been subtracted. The vertical patterns correspond to different crystal man-
ufacturers...... 173
xxvii A.2 Deviation from mean time after cluster timing correction...... 174 A.3 π0 peaks for signal candidate in data when combined with another
photon in the event. All π0 (blue), π0 candidates which are rejected by a 2.5σ timing cut (red), and the subtracted distribution (green). . 174 A.4 Candidate π0 which are rejected by timing cuts by cuts on timing significance, σ ...... 175
B.1 Photon energy spectrum near 1 GeV with (red) and without (blue) the crystal edge correction turned on (left). Photon energy spectrum in CM frame for photons satisfying E 1GeV < 0.01GeV (right). 178 | γ,LAB − | B.2 Background subtracted energy spectrum after (left) and before (right)
the edge correction is turned on. The spectra are shown with non- peaking background subtracted (top) and all background subtracted (bottom)...... 178 B.3 Photon energy spectrum of clusters with lateral moment equal to zero,
and passing all analysis cuts...... 179 B.4 Fit to full data sample after background subtraction. The cyan curve
is the sum of the χb peaks, and the red curve is the ISR PDF. The
histograms are χb (cyan), ISR (red), ηb (blue), the extra peak (green) and their sum (black). The ISR yield is floating in the fit. The top
plot displays the fit where the width of the spurious peak is fixed to detector resolution and the bottom displays the fit in which the width is allowed to float...... 181 B.5 Smooth background subtracted fit of energy spectrum with no primary
vertex correction...... 182
xxviii Chapter 1
Introduction
A fundamental test of the properties of Quantum Chromodynamics (QCD) is pro- vided by measurements of the properties of bound states of quark-antiquark pairs.
Calculations in QCD are extremely complicated but in certain limits the equations of QCD simplify. In the limit of high momentum transfers, the coupling between the quark and antiquark is weak, a limit in which calculations based on perturbation theory are tractable. The nonrelativistic limit also provides additional simplifications because in this limit relativistic degrees of freedom are removed. States consisting of
a bottom quark and antiquark pair provide a system in which both of these limits apply, and are therefore particularly interesting. While many states consisting of b¯b pairs were discovered decades ago, the ground state of the b¯b system, referred to as
the ηb meson, has remained elusive. This dissertation presents an analysis of data
from the BABAR experiment that provides the first observation of the ηb meson.
1 2 CHAPTER 1. INTRODUCTION
1.1 The Standard Model
Electrodynamics describes the interactions of electrically charged particles. A theory for properly treating electrodynamics within a quantum framework, quantum electro- dynamics (QED), arose in the 1920s from the study of the the electromagnetic field in quantum mechanics. As well as subsuming classical electrodynamic phenomena, QED predicts lepton magnetic moments and small deviations from classical behavior in atoms. In 1925, Born, Heisenberg and Jordan took the first step by expanding the electromagnetic field in terms of a set of independent harmonic oscillators [1]. Dirac, in 1927, extended this work with a fully quantum mechanically treatment of the electromagnetic field [2].
The leading-order cross sections for many processes (i.e.; pair annihilation[3], bremsstrahlung [4], Bhabha scattering [5]) were successfully calculated in the 1930s. However, the calculation of higher-order corrections were fraught with technical prob-
lems related to divergences. In the late 1940s, Feynman, Schwinger, and Tomonaga discovered how to absorb these divergences into “renormalized” masses, charges and field strength operators, a process known as renormalization [6, 7, 8].
Eventually it became understood that the criterion that the infinities of a theory can be absorbed in this way is useful in classifying theories. Indeed, renormalizabil- ity is a central feature of the theories of the weak and strong interactions. In the
1960s, Glashow, Weinberg and Salam proposed the SU(2) U(1) theory of the weak ⊗ interactions [9], in which the renormalizability of the theory plays a crucial role. The strong interaction, described further below, is a renormalizable gauge theory based on the SU(3)c gauge group. These theories form collectively what is now referred to as the Standard Model of particle physics. 1.2. QUANTUM CHROMODYNAMICS 3
1.2 Quantum Chromodynamics
1.2.1 Historical Overview
In November 1974, a narrow resonance with a large mass ( 3095 GeV/c2) was discov- ∼ ered simultaneously at Brookhaven and SLAC [10]. The discovery of the J/ψ, as it is
now known, was puzzling as the J/ψ is exceptionally narrow for a meson of that mass. The SLAC measurements determined that the J/ψ decays primarily to hadrons and therefore decays via the strong interaction. Because there is a large amount of phase space for the decay, the J/ψ should have a short lifetime and decay very rapidly via
the strong interaction, leading to a large total width for the J/ψ resonance. The solution to the puzzle lay in the fact that the J/ψ was the first example of a bound state of c and c¯ quarks to be discovered. Bound cc¯ states are referred to generically as charmonium, after the corresponding e+e− bound state, positronium. The ψ0, the
first excited state of the bound cc¯ system, was discovered a couple weeks later. A few years later, in 1977, the corresponding bound state of b and ¯b quarks, the Υ , was discovered [16]; these states are known collectively as bottomonium.
The long lifetime of the J/ψ led to the realization that Quantum Chromodynamics (QCD) provided the correct description of strong interaction phenomena. QCD is a theory of spin-1/2 particles, called quarks, which carry a charge (called “color” charge)
under the non-Abelian group SU(3)c. Interactions between particles carrying color charges are mediated by spin-1 particles called gluons. The Lagrangian describing QCD [17] is N f 1 = q¯ (iγµD m )q G Gµν L k µ − k k − 4 µν Xk=1 where D = (∂ ig taAa ) is the gauge covariant derivative, Aa is the gluon field, µ µ − s µ µ 4 CHAPTER 1. INTRODUCTION
a a Gµν is the gluon field strength, mk are the quark masses, t are the generators of the
fundamental representation of SU(3) and the qk are the quark fields. This theory, combined with the SU(2) U(1) electroweak gauge theory, represents the Standard W ⊗ Y Model of particle physics, which describes all terrestrial particle physics phenomena.
The QCD coupling constant (gs) is smaller in processes involving high momentum transfer, a property known as asymptotic freedom [11, 12], described further below. This property implies that at high energy, the coupling constant is small enough that it is possible to use perturbation theory to describe QCD. As a result of the discovery of asymptotic freedom, for the first time, it was possible to quantitatively
describe strong interaction phenomena [14, 15]. For simplicity, we will often refer
2 to the QCD analog of the QED fine structure constant, αs = gs /(4π) as the QCD coupling constant.
1.2.2 Asymptotic Freedom
Calculations of Feynman diagrams in QCD are plagued with divergences resulting
from integrations over the momenta of all virtual particles. In order to create a mathematically well-defined theory, these integrals in QCD must be cut off at some energy µ, in a process known as regularization. Imposing the requirement that the final results do not depend upon the choice of µ leads to the requirement that the strength of the QCD coupling changes (or “runs”) with energy, determined by the
beta-function, β(αs).
The beta-function can be calculated in perturbation theory and is given by
dα α α 2 µ s β(α ) = 2α β s + β s + , (1.1) dµ ≡ s − s 0 4π 1 4π · · · 1.3. THEORETICAL APPROACHES TO QCD 5
where
11 4 β = C T n 0 3 A − 3 F f 34 20 β = C2 C T n 4C T n , . . . , 1 3 A − 3 A F f − F F f
and C = N = 3, C = (N 2 1)/(2N ) = 4/3 and T = 1/2. A c F c − c F
Because of the number of colors (Nc) and quark flavors (nf ) observed in nature, the β function of QCD is negative, leading to a weakening of the QCD coupling at high
energies (or small spatial separations). Conversely, at low energies (or large spatial separations), the QCD coupling becomes large, and the potential energy between charges diverges, causing color charges (the QCD analog of electric charges) to be permanently confined. This explains why free quarks have never been observed,
although they interact as nearly free particles in high energy deep inelastic scattering experiments.
The unexpected long life of the J/ψ can be explained by the asymptotic freedom of QCD. If this is the case, the gluons must carry the four-momentum of the decaying
particle, a large momentum, leading to a small value of αs. The small value of the strong coupling constant in turn suppresses the decay rate.
1.3 Theoretical Approaches to QCD
1.3.1 Quark Potential Models
The equations of QCD are exceptionally difficult to solve. Due to the large value of
αs at low energy, perturbation theory, the primary tool for solving most problems in 6 CHAPTER 1. INTRODUCTION
quantum field theory, is no longer applicable. As a result, for many problems in QCD, one is forced to use phenomenological models of the potential energy between a quark and an antiquark, known as potential models, to make predictions involving strong interaction phenomena. Originally, quark potential models were the only tool for understanding QCD phenomena, but they are slowly being replaced by calculations rigorously derived from QCD.
In potential models, one constructs an quark-antiquark potential with enough free parameters to fit the spectrum observed by experiment. At short distances, due to asymptotic freedom, the strong coupling constant is small and so single gluon exchange dominates. This gives rise to a Coulombic form for the potential, analogous to that arising in QED from single photon exchange. At longer distances, color charges are confined, which leads to the requirement that the potential energy between colored particles must diverge at long distances. A popular form of the potential, the so-called “Cornell potential”, is [18] 4 α (r) V (r) = s + σr −3 r where σ 0.18 GeV2 and gives qualitative agreement with the experimental spec- ≈ trum. A plot of this potential is given in Fig. 1.1. To this potential, spin-dependent terms describing spin-spin, spin-orbit and tensor interactions, similar to those de- scribing the hydrogen atom, are added. Recently, progress has been made in deriving the quark-antiquark potential from perturbative QCD [20].
1.3.2 Lattice QCD
For calculations involving processes at low energies the strong coupling constant be- comes of order one, rendering perturbative expansions in αs divergent. An alternate, 1.3. THEORETICAL APPROACHES TO QCD 7
2500 2000 1500
V(r) (MeV) 1000 500 0 -500 -1000 -1500 -2000 -2500 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 r (fm)
Figure 1.1: Cornell model potential for σ = 0.18 GeV2. non-perturbative description of QCD, pioneered by Wilson, describes gauge fields in terms of closed loops on a discrete space-time lattice [21]. This description of QCD has the advantage of regularizing QCD in a way that manifestly preserves gauge invariance.
In lattice QCD, observables are calculated by numerically evaluating path integrals in Euclidean space-time by taking expectation values over a large number of gauge field configurations. Expectation values of an operator (φ) are calculated using the O equation dφ [φ] exp ( S[φ]) < >= O − . (1.2) O dφ exp ( S[φ]) R − where S[φ] is the action describing theR φ field and the denominator is a normal-
ization factor. This integral is calculated numerically by averaging over gauge field configurations, using a procedure which ensures the proper exp ( S[φ]) weighting. − Statistical errors on these integrals can, in principle, be made arbitrarily small. 8 CHAPTER 1. INTRODUCTION
However, lattice QCD is extremely computationally intensive, limiting the precision with which it is practical to carry out these computations.
Until recently, most lattice simulations ignored light quark vacuum polarization (the so-called “quenched” approximation), which is computationally expensive to simulate. This is an uncontrolled approximation with errors which are difficult to es- timate. However, recent results, which include light quark vacuum polarization, have
succeeded in providing very precise results which are in agreement with experiment for a wide range of observables, including the heavy quarkonium spectrum [24] shown in Fig. 1.2, calculated by the UKQCD and HPQCD collaborations.
10.6
10.4 3S Mass (GeV) 2P 10.2 1D
10 2S 1P 9.8
Experiment 9.6 Quenched Unquenched 1S 9.4
Figure 1.2: Spectrum of b¯b levels from a recent lattice prediction [24]. Closed and open symbols are from coarse (lattice spacing a 0.12 fm) and fine (a 0.09 fm) ≈ ≈ lattices, respectively. Squares and triangles denote unquenched and quenched results, respectively. Lines represent experiment.
Lattice calculations face several sources of systematic error, in addition to the errors discussed above. Calculations on any finite lattice can only approximate con- tinuum QCD, and an extrapolation of results obtained at finite lattice spacing to zero 1.3. THEORETICAL APPROACHES TO QCD 9
lattice spacing is always necessary. This is very computationally intensive as, for a given volume, the computational requirements naively scale as the fourth power of
the inverse lattice spacing, and in practice the scaling is often worse. To reduce the systematic error due to discretization, smaller lattice spacings can be used in combi- nation with a smaller simulation volume but, of course, the volume must fully contain the hadron of interest, necessitating a trade-off.
An additional source of uncertainty is due to the fact that integrals involving light quark vacuum polarization becomes noisier as the light quark masses are de- creased, requiring higher simulation statistics to achieve an identical statistical error
on the path integral. Simulations are performed with unphysically large light quark masses, and an extrapolation to physical light quark masses is necessary, incurring an additional systematic error.
In lattice simulations, full Lorentz symmetry is broken down to discrete symme- tries, often creating artifacts which should disappear in the continuum limit. Lattice simulations require an understanding of how these lattice artifacts can arise. As the discretization of the action is not unique, these errors can be studied by altering the discretization.
In B meson physics, lattice techniques are often used to predict observables that are relevant in searches for physics beyond the Standard Model. The bottomonium spectrum gives a large number of experimentally well-known quantities which should be straightforward to predict using lattice techniques. Therefore, aside from the intrinsic interest in the prediction of the bottomonium spectrum, the agreement of the lattice QCD predictions with experiment constitutes an important validation of
lattice techniques. 10 CHAPTER 1. INTRODUCTION
1.3.3 Non-relativistic quantum chromodynamics
An important tool in the analysis of quantum field theories involving multiple energy scales is the effective field theory. An effective field theory (EFT) is obtained by integrating out degrees of freedom above a given energy scale, resulting in a simpler, but equivalent, theory in which calculations are more manageable.
Non-relativistic QED (NRQED) was the first EFT for treating nonrelativistic bound states [30]. NRQED provides an ideal framework for the description of positro- nium (the bound state of an electron and positron) and muonium (a neutral eµ state) and greatly facilitates the calculation of higher order corrections. Some recent successes are the calculation to (α∈ ln α) corrections to the hyperfine splitting in O positronium (for a precision of 2.0 ppm) and the most precise value of mµ/me, which provides the best determination of the mass of the muon [32].
The successes of NRQED led to the development of its analogue in QCD, non- relativistic QCD [31]. Nonrelativistic quantum chromodynamics (NRQCD) is the
effective field theory resulting from integrating out modes of energy and momentum describing heavy quark-antiquark pairs from QCD Green’s functions. Heavy quarko- nium are characterized by three separate scales, often referred to as the hard, soft, and ultrasoft scales. The hard scale is associated with the mass m of the heavy quark.
The soft scale is determined by the relative momenta of the quark and anti-quark, mv. Finally, the ultrasoft scale is the kinetic energy E = mv2 of the heavy quark and anti-quark. All of these scales are assumed to be greater than the characteristic scale
of QCD, ΛQCD.
In the non-relativistic limit v c heavy quark-antiquark production is suppressed, 1.3. THEORETICAL APPROACHES TO QCD 11
decoupling the heavy quark and antiquark, and allowing a projection onto states con- taining exactly one heavy quark and antiquark. It is therefore sufficient to use Pauli spinors to describe the heavy quark and antiquark separately. NRQCD has obtained considerable successes, such as the quarkonium production rate at the Tevatron [34] and removing the infrared divergences in the calculation of the decay rates of P -wave quarkonium [33].
The NRQCD Lagrangian is an expansion of the form
O (µ) L = c (α (m), µ) n , (1.3) k s mk Xk
where the coefficients ck, called Wilson coefficients or matching coefficients, are deter- mined by equating the Green’s functions of NRQCD and full QCD at the matching scale, µ. For example, the part of the NRQCD Lagrangian that is bilinear in the quark fields is, to order (m v4), O b
L † 1 2 1 4 cF = ψ iD0 + D + 3 D + σ gB 2mQ 8mQ 2mQ ·
cD cS + 2 (D gE gE E) + i 2 σ (D gE gE D) ψ. (1.4) 8mQ · − · 8mQ · × − × !
The third through sixth terms are the QCD analog of terms familiar from the
analysis of the hydrogen atom in QED: the relativistic correction, the hyperfine term, the Darwin term and the spin-orbit coupling, respectively.
A lattice simulation must contain the entire hadron within the simulation volume but the lattice spacing must remain fine enough to resolve spatial details and the
decay in time of correlation functions, necessitating a large number of lattice points. 12 CHAPTER 1. INTRODUCTION
The largest momentum that can be simulated on a discrete lattice is π/L = π/(aN), whereas the smallest simulated momentum is π/a. As a result, the reduced range of
scales in NRQCD makes it ideal for lattice simulation.
1.3.4 Potential non-relativistic quantum chromodynamics
NRQCD contains additional degrees of freedom of order mv, as well as mv2. Potential non-relativistic quantum chromodynamics (pNRQCD) is derived from NRQCD by in- tegrating out these degrees of freedom of order mv [35]. In this theory the matching
coefficients between NRQCD and pNRQCD take the form of quark-antiquark poten- tials.
In this EFT the theory is simplified to an extent that, in some situations, analytic calculations of the spectrum are possible in a manner identical to perturbation theory with a Schr¨odinger-like potential. The potential used in potential model approaches to QCD arises naturally as a Wilson coefficient in the pNRQCD approach.
The pNRQCD Lagrangian has the schematic form
O (µ) L = c (α (m), µ) rnV (r, µ) n (1.5) k s mk Xk where V (r, µ) are matching coefficients, r is the quark-antiquark separation, and µ is the renormalization scale. As an example of the sort of terms which arise in the expansion, the hyperfine splitting in the bottomonium ground state is given by
(2) 4CF π 2 σ1 + σ2 Hspin = DS2,s 2 S , S = (1.6) 3mb 2
(2) where σ1 and σ2 are the spin operators of the quark and anti-quark, and DS2,s is the 1.4. BOTTOMONIUM 13
expectation value of the operator
2 2 (2) 4πCF DS ,s (3) V 2 (r) = δ (r). (1.7) S ,s 3
1.4 Bottomonium
At short distances QCD is dominated by one-gluon exchange leading to a spectrum which is similar to the Coulomb spectrum. There are significant corrections to the
Coulombic potential due to non-perturbative physics. The current picture of the bottomonium spectrum below BB¯ threshold is shown on Fig. 1.3, where the numerous electric dipole transitions are omitted. Symmetries under charge conjugation and parity impose selection rules on the allowed transitions.
1.5 Electromagnetic transitions
Electromagnetic transitions between bottomonium states are usually treated by po- tential models in a Schr¨odinger formalism, analogous to the treatment of transitions in atoms. Though transition rates can, in principle, be computed with lattice meth- ods, only one group has attempted calculations of these rates, and only for transitions between charmonium states [37].
The lowest order electromagnetic transitions are the electric dipole (E1) and mag- netic dipole (M1) transitions. Electric dipole transitions in quarkonium have ∆` = 1 and ∆s = 0 whereas the magnetic dipole transitions have ∆` = 0 and ∆s = 1. This results in a selection rule that, for dipole transitions, the parity of the state changes in a electric dipole transition but not in a magnetic dipole transition. 14 CHAPTER 1. INTRODUCTION
Figure 1.3: Spectrum of b¯b levels and possible transitions between levels. Figure from [36].
Electric dipole transitions can be calculated in a potential model using
4 Γ = e2 αC (2J + 1)E3 n L r n L 2 (1.8) E1 3 Q if f γ |h f f | | i ii|
where Cif is a statistical factor, equal to 1/9 for transitions between S and P states,
eQ is the charge of the quark, Jf is the total angular momentum of the final state,
Eγ is the photon energy, r is the interquark separation, and the ni,f , Li,f are the quantum numbers of the initial and final states.
Magnetic dipole transitions can be calculated in a potential model using
4 Γ = e2 αC (2J + 1)E3 n L r n L 2 (1.9) M1 3 Q if f γ |h f f | | i ii| 1.6. THE GROUND STATE OF BOTTOMONIUM, THE ηB MESON 15
where Cif is a statistical factor, equal to 1 for ∆` = 0 transitions.
The calculation of magnetic dipole transition rates in potential models is compli- cated due to several factors. In potential models the Lorentz structure of the confining potential (vector or scalar) is an arbitrary choice. Also, there is a dependence upon the anomalous magnetic moment of the quark. Finally, relativistic corrections contain an explicit dependence on the quark masses, which in potential models are simply fit parameters rather than rigorously defined quantities.
Magnetic dipole transitions can also be calculated in a model-independent fashion in pNRQCD [38]. However, the calculations performed to date are only applicable
2 to the weak-coupling regime (mv & ΛQCD), a condition which is not satisfied by the excited heavy quarkonium states.
All of these approaches assume that the photon energy is small enough that a
multipole expansion can be used to describe the electromagnetic field. The multipole expansion converges for kr 1, a condition which may not be satisfied by the excited bottomonium states. Recently, another approach has been proposed in which avoids the multipole expansion by factorizing the transition amplitude into a hard scattering
amplitude and non-perturbative wavefunction [39], analogous to the factorization of the pion form factor at large momentum transfer [40].
1.6 The ground state of bottomonium, the ηb me-
son
The Υ (1S) meson, the first bb bound state to be discovered, was found over 30 years ago yet until this study its pseudoscalar partner, the ηb meson, had still not been 16 CHAPTER 1. INTRODUCTION
discovered. The ηb meson, the ground state of bottomonium, is interesting because its dynamics are expected to be largely perturbative, due to its small spatial extent.
A measurement of the mass and width of the ηb meson would help to test lattice NRQCD, pNRQCD, and quark potential models in a regime where all three should be valid.
The expectations for the mass splitting between the ηb and Υ (1S) mesons vary from 36 to 100 MeV. A recent lattice calculation within the NRQCD framework by the HPQCD and UKQCD collaborations [24], including vacuum polarization for u, d, ands quarks, gives a prediction of 61(4)(12)(6) MeV/c2. The results are shown in Fig. 1.4. The errors are statistical/fitting and discretization errors; radiative cor- rections and relativistic corrections, respectively. In this calculation, the NRQCD matching coefficients are taken at lowest order in the expansion (tree level), and only operators up to order (α5m ) are included. A similar unquenched lattice O s b NRQCD calculation, done with a different fermion and gauge discretization (domain wall fermions and the Iwasaki gauge action [26, 27] at a single lattice spacing of 0.11 fm gives a result of 52.5 1.5 MeV/c2, where the errors are statistical only, ≈ and no continuum extrapolation is performed [28]. This hyperfine splitting is consis- tent with that obtained by the HPQCD and UKQCD collaborations on their 0.12 fm lattice. Another calculation of the hyperfine splitting was performed by the Fermilab
Lattice and MILC Collaborations [29] using the Fermilab action predicts a splitting of 54.0 12.4+1.2, where the second error results from the conversion from lattice −0.0 units to physical units.
The hyperfine splitting may also be calculated within perturbative QCD. Fig-
ure 1.5 shows the variation of the hyperfine splitting versus the renormalization 1.6. THE GROUND STATE OF BOTTOMONIUM, THE ηB MESON 17
scale at various levels of accuracy in the perturbative expansion. This calcula- tion, performed in the pNRQCD framework, results in a hyperfine splitting of 41 +8 11(th)−9(αs) MeV [19]. A recent calculation of the hyperfine splitting using the static quark potential (the quark-antiquark potential in the limit m ) from perturba- b → ∞ tive QCD to (α5m ) gives m m gives 44 11 MeV [22]. Theoretical calculations O s b Υ − ηb which are rigorously based on QCD (not potential models) are summarized in Ta- ble 1.1.
Table 1.1: Predictions for the hyperfine splitting mΥ (1S) mηb from lattice QCD and perturbative QCD calculations. −
Calculation Hyperfine Splitting ( MeV/c2) Reference +8 pNRQCD 41 11(th)−9(αs) [19] Pertubative QCD 44 11 [22] Lattice NRQCD 61 4 12 6 [24] Lattice NRQCD 52.5 1.5 [28] Lattice NRQCD 54.0 12 .4+1.2 [29] −0.0
Non-perturbative corrections arise at order (m (Λ3 /m3)) m α6 5 MeV [23]. O b QCD b ≈ b s ≈ The leading non-perturbative corrections are dominated by the vacuum expectation value of gluon pairs, the gluon condensate,
1 G2 = 0 α G Gµν 0 (1.10) h i 4h | s µν | i which shifts the hyperfine splitting by
2 301066767 π αsG LO ∆Ehf = h 4 6 iEhf (1.11) 92480000 mb αs
LO 4 4 where Ehf = CF αs(µ)mb/3 is the hyperfine splitting at leading order in αs. 18 CHAPTER 1. INTRODUCTION
Figure 1.4: Spectrum of b¯b levels from a recent lattice prediction [24]. Closed and open symbols are from coarse (lattice spacing a 0.12 fm) and fine (a 0.09 fm) lattices, respectively. Squares and triangles denote≈unquenched and quenched≈ results, respectively. Lines represent experiment.
The ηb meson is expected to decay almost exclusively to two gluons with a full width of around 10 MeV. The width of the ηb meson is expected to be smaller than that of the ηc due to a smaller value of the strong coupling constant at the b quark mass scale relative to the c quark mass scale (α (m ) 0.2 versus α (m ) 0.3). s b ≈ s c ≈
Predictions for the full width of the ηb are uncommon in the literature; rather, the di-photon width, the calculation of which is not beset with complications due to QCD radiative corrections, is usually computed. Assuming Γ Γ(η gg) and neglecting ≈ b → QCD radiative corrections, then the two-photon and two-gluon processes are equiva-
lent up to a vertex factor. The full width can then be calculated by multiplying the 1.6. THE GROUND STATE OF BOTTOMONIUM, THE ηB MESON 19
50 45
(MeV) 40 hfs
E 35 30 25 20 15 10 5 0 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5 ν (GeV)
Figure 1.5: Hyperfine splitting M M as a function of renormalization scale µ Υ (1S) − ηb in leading order (dotted line), next to leading order (dashed line), leading log (dot- dashed line), and next-to-leading log (solid line) approximations. For the next-to- leading log result, the yellow band corresponds to a variation α (M ) = 0.118 0.003. s Z Figure from [19]. two-photon partial width by
2 α (m )2 α (m ) 9 s b 1 + 7.8 s b 2.4 104. (1.12) e4α2 π ≈ × b where α and αs are the electromagnetic and strong coupling constants, respectively, and e is the b-quark mass in units of e [45]. The factor 2/9 is a color factor and b | | the QCD radiative correction is the quantity in brackets. Table 1.2, summarized from [41], shows predicted widths for the ηb.
It has been suggested by Sanchis-Lozano that mixing with a light Higgs could modify the predictions [42] for both the hyperfine splitting and the width of the ηb.
Until this study, the ηb meson had not been observed. The large amount of phase 20 CHAPTER 1. INTRODUCTION
Table 1.2: Predictions for the two-photon partial decay width, taken from [41], and the total width determined by scaling the two-photon decay width using Eq. 1.12.
Author Γ(η γγ) (keV) Γ (MeV) b → tot Kim [46] 0.384 0.047 9.2 1.1 Munz¨ [47] 0.22 0.04 5.3 1.0 Chao [48] 0.46 11.0 Ebert [49] 0.35 8.4 Fabiano [50] 0.47 0.10 11.3 2.4 Gupta [51] 0.46 11.0 +0.019 Penin [52] 0.659 0.089(th.)−0.018(δαs) 0.015(exp.) 15.8 2.2 Laverty [53] 0.39, 0.42 9.4, 10.1
space available to ηb decays leads to extensive fragmentation of the gg decay, leading to few preferred exclusive final states. The only experimental information about the decay of bottomonium states to exclusive hadronic final states is provided by
the CLEO experiment, which has measured exclusive decays of χbJ (nP ), J = 0, 2, n = 1, 2 states [43]. No single exclusive hadronic decay mode has a branching fraction
−3 of greater than 10 . The χbJ (nP ), J = 0, 2 states decay via two gluons (as the ηb)
so their decays should be qualitatively similar to those of the ηb.
A promising way to produce and detect the ηb meson is via the photon produced
in magnetic dipole transitions from the Υ (3S) to the ηb. The rate can be calculated somewhat crudely in potential models. In the nonrelativistic approximation and including finite size corrections, the rate is given by
2 4 eb 2 3 Γ(Υ (3S) ηbγ) = α 2 I k (1.13) → 3 mb
where α is the fine structure constant, e is the b quark charge in units of e , k is the b | |
photon momentum, and mb is the b quark mass. The overlap integral I is defined by 1.6. THE GROUND STATE OF BOTTOMONIUM, THE ηB MESON 21
I = f j (kr/2) i (1.14) h | 0 | i where j0(x) is the spherical Bessel function of zeroth order. The spherical Bessel function takes into account the correction for the finite size of the bottomonium states [59], neglected in Eq. 1.9.
The transitions Υ (n0S) γη (nS) are conventionally separated into two types → b 0 0 based on the radial quantum numbers of the Υ (n S) and the ηb(nS): “allowed” (n = n) and “hindered” n0 = n. Since kr << 1, expanding j (kr/2) = 1 + k2r2/24 + 6 0 · · · in Eq. 1.14, and using the orthogonality of the wavefunctions, we see that the rate of hindered transitions is given entirely by higher-order and relativistic corrections. Therefore the predicted rate is subject to a substantial theoretical uncertainty. The expected branching fraction for the transition Υ (3S) γη ranges from 0.5 10−6 → b × to a few times 10−3 with most predictions around 10−3. Godfrey and Rosner give a recent review [60] of branching fractions and hyperfine splittings.
Figure 1.6, reproduced from the paper describing the ηb search by the CLEO experiment, shows the from the CLEO experiment and several predictions for the branching fraction Υ (3S) γη . → b
ALEPH [54], L3 [55], and DELPHI [56] have searched for the ηb meson in two- photon collisions, and provided 95% confidence level upper limits on the product
Γ (η ) BR(η ) for various assumed decay modes [57]. The results are summa- γγ b × b rized in Table 1.3, which is reproduced from the report by the Heavy Quarkonium Working Group [57]. The best limits are from the ALEPH experiment. Assum-
ing an ηb two-photon decay width of 660 MeV, the ALEPH limits correspond to (η 4 charged) < 0.07 and (η 6 charged) < 0.20 at the 95% confidence level. B b → B b → 22 CHAPTER 1. INTRODUCTION
: 2 Hyperfine Splitting MΥ(1S) - Mηb(1S) (MeV/c ) 30 50 70 90 110 130 150 3 | | | | | | | | | | | | | , Zambetakis,Byers 83 -3 , 2 Godfrey-Isgur 85 B
, Godfrey-Isgur 85 A
1
Branching Ratio in units of 10 90% CL UL CLEO-III
, Lahde,Nyfalt,Riska 99 A
0 880 900 920 940 960 980 1000 Eγ (MeV)
Figure 1.6: Theoretical predictions for Υ (3S) γηb branching fraction and CLEO upper limit. →
At hadron colliders, it is very difficult to trigger on ηb decays into hadrons; instead,
one must trigger on rare decays of the ηb. In an unpublished analysis, CDF also has searched for the η , in the decay mode η J/ψJ/ψ. Three candidates are observed b b → compared to an expectation of 3.8 events [58].
1.7 Conclusion
Measurements of the properties of the ηb meson play an important role in the study
of nonrelativistic bound states in QCD. In this dissertation we observe the ηb meson 1.7. CONCLUSION 23
Table 1.3: Limits (95% confidence level) on ηb two photon partial width times branch- ing fraction from LEP.
Experiment final state Γ B (keV) Reference γγ × ALEPH 4 charged < 0.048 [54] 6 charged < 0.132 [54] L3 K+K−π0 < 2.83 [55] 4 charged < 0.21 [55] 4 charged π0 < 0.50 [55] 6 charged < 0.33 [55] 6 charged π0 < 5.50 [55] π+π−η0 < 3.00 [55] DELPHI 4 charged < 0.093 [56] 6 charged < 0.270 [56] 8 charged < 0.780 [56]
for the first time, and measure relevant properties of the ηb. These properties include measurements of hindered magnetic dipole transition rates to the ηb, and the ηb mass, from which the hyperfine splitting ∆m = m m is obtained. From the Υ (1S) − ηb hyperfine splitting and an analytic calculation of the splitting [19], we obtain the strong coupling constant, αs, in a unique way. Chapter 2
The BABAR Detector
2.1 Physics Motivation
2.1.1 CP violation
A curious fact about our universe is that it contains matter. The conventional measure of the baryon content of the universe is the ratio of the number density of baryons to that of photons. This baryon-to-photon density ratio, as measured by the WMAP cosmic microwave background experiment, is η = n /n = (6.225 0.170) 10−10 [61]. b γ × The predominance of matter can be explained by out-of-equilibrium CP -violating reactions which violate baryon number in the early universe, the so-called Sakharov
mechanism [62]. The Standard Model contains all of the necessary ingredients to qualitatively explain the existence of matter in the universe; however, the quantitative prediction of the baryon-to-photon ratio in the Standard Model is much smaller than observed [63]. This leads to the suggestion that perhaps there are additional sources
of CP -violation in addition to those found within the Standard Model.
24 2.1. PHYSICS MOTIVATION 25
One of the primary motivations for the BABAR experiment is the study of CP - violation, through the time dependence of neutral B meson decays, in order to deter- mine if there are sources CP -violation beyond those found in the Standard Model. The decay of the Υ (4S) resonance produces an entangled B0 B¯0 state. The B0 and − B¯0 mesons can oscillate into each other via box diagrams involving the W boson and u, c, t quarks. Before one of the mesons decays, there must be exactly one meson and one anti-meson; once either meson decays, this coherence is lost, and the other meson can oscillate freely before decaying. The measurement of the time dependence of the resulting particle-antiparticle asymmetries can be used to measure CP -violation parameters.
To achieve this goal it is necessary to determine the separation between the decay vertices of the two mesons, a task that is complicated by the B mesons being produced almost at rest (p 330 MeV) in the Υ (4S) frame. Since the lifetime of the B0 B ≈ −12 0 meson is τ 0 = (1.530 0.009) 10 seconds and the B meson is produced nearly B × at rest in the frame of the Υ (4S), the B0 meson travels only d = γvτ 30µm before ≈ decaying. With current technology this separation is too small to be measured with high enough accuracy. By constructing a collider in which the CM frame moves at relativistic speeds relative to the lab frame, as is done at the PEP-II-BABAR B-factory, the separation between the two decay vertices is improved.
For studies of CP -violation, it is critical to have a excellent vertex detector for very precise measurements of the position of the two decay vertices. However, it is also necessary to have general purpose detector with high reconstruction efficiency of neutral clusters and charged tracks, making the BABAR detector an excellent detector for bottomonium physics, as well as CP -violation physics. 26 CHAPTER 2. THE BABAR DETECTOR
2.1.2 Bottomonium Physics
Though the Υ (4S) resonance is ideal for studying B-meson decays, the transition
rates to other bottomonium resonances are overwhelmed by the rate of Υ (4S) decays to B B¯ pairs. The lower Υ resonances are better suited for studying bottomonium − transitions because they are below threshold for B B¯ production. −
2.2 PEP-II
The PEP-II e+e− storage rings were originally designed to operate at a center of
mass energy √s = 10.58 GeV, the peak of the Υ (4S) resonance. To support the primary purpose of the BABAR experiment, the center of mass is boosted relative to the laboratory frame by βγ = 0.56 by using an electron beam energy of 9.0 GeV and a positron beam energy of 3.1 GeV. This increases the average separation of the decay
vertices of B mesons resulting from Υ (4S) decays to roughly < βγcτ 0 > 260µm. B ≈ For data taking at the peak of lower resonances, the energy of positron beam is unchanged while the energy of the electron beam is reduced to 8.61 GeV and 8.07 GeV for the Υ (3S) and Υ (2S) resonances, respectively.
The direction of the electron beam and positron beam are referred to as the “forward direction” and “backward direction”, respectively.
2.3 The BABAR Detector
The BABAR detector, described in detail in [64] and pictured in Fig. 2.1, is composed
of five principal active subdetectors: 2.4. SILICON VERTEX TRACKER 27
Silicon Vertex Tracker (SVT) - provides dE/dx measurements and precise mea- • surements of track vertex parameters and is the only source of tracking for low
momentum tracks
Drift Chamber (DCH) - in combination with the SVT, measures momenta of • tracks and provides dE/dx measurements for particle ID
Calorimeter (EMC) - measures energies of electron and photon showers; also • used for electron and neutral hadron ID
Detector of Internally Reflected Light (DIRC) - provides particle identification, • particularly π/K/p separation
Instrumented Flux Return (IFR) - provides muon and K0 identification • L
2.4 Silicon Vertex Tracker
Trajectories of charged particles very close to the interaction point are measured using the Silicon Vertex Tracker (SVT). The SVT, shown in Fig. 2.2, consists of five layers of double-sided silicon strip detector in which the strips in the two sides are arranged perpendicular to each other, to provide z-position information. The close proximity of the silicon vertex tracker to the interaction point allows for an improved determination of the primary vertex of a physics event, relative to the precision determined in the absence of the silicon vertex tracker. While precision of the silicon vertex tracker is imperative in the the studies of time-dependent CP violation, it is less important in the study of bottomonium physics. 28 CHAPTER 2. THE BABAR DETECTOR
Detector C L Instrumented Flux Return (IFR)) 0 Scale 4m I. P . Barrel Superconducting BABAR Coordinate System Coil y 1 0 1 5 1 7 4 9 x Electromagnetic Cryogenic 1149 4050 1149 Calorimeter (EMC) z Chimney 3 7 0 Drift Chamber (DCH) Cherenkov Detector Silicon Vertex (DIRC) Tracker (SVT)
IFR Magnetic Shield 1 2 2 5 Endcap for DIRC Forward 3 0 4 5 End Plug Bucking Coil 1 3 7 5 Support Tube 8 1 0 e– e+
Q 4 Q 2
Q1 3 5 0 0 B1
Floor 3-2001 8583A50
Figure 2.1: BABAR detector longitudinal section.
The SVT complements the measurements of the drift chamber, a description of which appears in the following section. The drift chamber cannot adequately measure low momentum tracks (below about 120 MeV/c) and for these tracks the SVT must fully determine track parameters. The high precision measurements in the SVT are especially important for tracks with high momentum, as these have a small sagitta in the drift chamber. These high precision measurements are also critical for deriving an accurate Cherenkov angle in the DIRC.
The inner three layers of the SVT are necessary for accurately measuring decay vertices and are placed as close to the beam pipe as possible to reduce errors due 2.4. SILICON VERTEX TRACKER 29
580 mm Space Frame
Bkwd. support cone 520 mrad Fwd. support3 50 mrad cone e- Front end e + electronics
Beam Pipe
Figure 2.2: Silicon vertex tracker longitudinal section. to multiple scattering. The outer two layers are used for extrapolating SVT tracks to the drift chamber. These two layers are arranged in an arch shaped geometry to maximize the crossing angle of particles near the edge of the SVT acceptance.
It is critical to maintain a minimum of material in the SVT. Otherwise, multiple scattering degrades momentum resolution for charged tracks. Also, photon conver- sions in the material degrade photon measurements in two ways: the resulting e+e− pairs lose significant energy by bremsstrahlung before reaching the calorimeter or the leptons may be lost completely due to bending in the magnetic field, leading to re- duced efficiency. As a result of its effect on photon efficiency, material further from the calorimeter is particularly harmful. At 90◦ the SVT is less than 0.04 radiation lengths thick.
The required resolution (about 80 µm) along the z-axis is determined by the requirement that the vertices of the B meson daughters of the Υ (4S) resonance are separately resolved, a capability necessary for time-dependent CP -violation studies. 30 CHAPTER 2. THE BABAR DETECTOR
Additionally, decay vertices of light charm mesons must be resolved, leading to a requirement of a resolution of 100µm in the x y plane. ≈ − Measurements of position and angle of tracks are dominated by the SVT measure- ments. The resolution of the helix parameters can be derived using cosmic rays that pass through the IP. An incident cosmic ray is reconstructed as two separate tracks and the difference between the reconstructed track parameters of the two tracks gives the resolution parameters. They are
σd0 = 23µm σφ0 = 0.43 mrad (2.1) σ = 29µm σ = 0.53 10−3 z0 tan λ × where d0, φ0, and z0 are the distance of closest approach, φ0 and z0 are the φ and z of the track at the distance of closest approach, and λ is the dip angle (the angle between the track momentum and the z-axis).
2.5 Drift Chamber
The principal device for measuring the trajectories of charged particles is a 40-layer drift chamber (DCH). The DCH also provides measurements of ionization loss dE/dx for particle ID, which are especially important for particle ID of low momentum tracks and tracks outside the acceptance of the DIRC. The DCH is also the only source of tracking for charged particles outside the acceptance of the SVT. This includes charged pions from KS decays, which may occur outside the SVT acceptance, due to the long K lifetime of 0.895 10−10s. In the central angular region of the detector this S × measurement complements the PID capabilities of the DIRC. In the extreme forward 2.5. DRIFT CHAMBER 31
and backward regions, where the DIRC is not present, and in the gaps between the DIRC bars in the barrel, the DCH is the only source of PID information.
The DCH is shown in Fig. 2.3. As charged particles traverse material they lose
630 1015 1749 68
Elec– tronics 809
485 27.4 1358 Be 17.2 236 e– 464 IP e+ 469
1-2001 8583A13
Figure 2.3: Drift chamber longitudinal section. Measurements are in mm.
energy by ionization, atomic excitation, and radiation. Hits are formed when charged particles ionize the molecules of the drift chamber gas and these electrons drift toward the anode. The drift chamber uses the drift velocity of electrons in a gas resulting from these ionizations to calculate the position of interactions of charged particles that propagate through the DCH. Hexagonal cells are used to obtain approximately circularly symmetric isochrones.
A diagram of the wire placement for the first four superlayers is shown in Fig. 2.4. Longitudinal position information is obtained by placing 24 of the 40 layers at a small stereo angle (50-76 mrad) with respect to the z-axis. The overall resolution is roughly 140 µm, as shown in Fig. 2.5. For each cell, the relation between the time and
distance of closest approach to the wire is calibrated using e+e− and µ+µ− events. 32 CHAPTER 2. THE BABAR DETECTOR
To avoid bias in this calibration, the cell of interest is excluded from the calculation of the distance of closest approach.
16 0 15 0 14 0 13 0
12 -57 11 -55 10 -54 9 -52
8 50 7 48 6 47 5 45
4 0 3 0 2 0 1 0 Layer Stereo
4 cm Sense Field Guard Clearing
1-2001 8583A14
Figure 2.4: Drift chamber wire placement for first four superlayers. Stereo angles are in mrad.
Information regarding dE/dx is obtained from the measured charge in each cell, corrected for several factors including global corrections for gas pressure and temper- ature, individual cell geometry, space charge buildup, non-linearities in energy loss at large dip angles, and variation of charge collection efficiency as a function of entrance angle. The dominant effects on the resolution come from global corrections. The overall dE/dx resolution is observed to be 7.5%, close to the expected performance of 7%. This provides excellent K p separation for momenta less than 1 GeV/c2, as − shown in Fig 2.6. 2.5. DRIFT CHAMBER 33
0.4
0.3
0.2 Resolution (mm) 0.1
0 ±10 ±5 0 5 10 1-2001 8583A19 Distance from Wire (mm)
Figure 2.5: Single cell resolution as a function of distance from sense wire.
Hits are reconstructed to form tracks in 3-space using a Kalman filter approach [65] which includes dE/dx measurements, the distribution of material in the detector and magnetic field inhomogeneities.
The transverse momentum resolution, σpt , is dominated by the measurements from the DCH. It it derived from cosmic ray events, as discussed in Section 2.4, and is p σ /p = (0.13 0.01)% t + (0.45 0.03)%. pt t × 1GeV where the first term is related to the measurement accuracy, and the second is due to multiple Coulomb scattering. Multiple scattering is seen to dominate momentum resolution for all but the highest momentum tracks. 34 CHAPTER 2. THE BABAR DETECTOR
104 d p
K dE/dx
103 π
e µ
10±1 1 10 1-2001 8583A20 Momentum (GeV/c)
Figure 2.6: Specific energy loss versus momentum. Lines show predictions for various mass hypotheses from the Bethe-Bloch equation.
2.6 Detector of Internally Reflected Light
In many analyses accurate particle identification is necessary both for background reduction and the decision of the appropriate mass hypotheses to be used in vertexing. The Detector of Internally Reflected Light (DIRC) is shown schematically in Fig 2.7.
The study of CP violation requires the ability to tag the flavor of one of the B mesons using kaons. The kaons produced in B decays have high momentum, generally above 1 GeV/c. When particles travel faster than the speed of light in a material, a
cone of light, referred to as a Cherenkov cone, is produced at an angle θC with respect
to the flight direction, where cos θC = 1/(nβ), n is the refractive index of the medium and β = v/c. The particle ID system used in BABAR is a new type of Cherenkov detector in which the Cherenkov ring is imaged by propagating the Cherenkov light created in a quartz bar with a 1 2 cm2 cross section through the bar by internal × reflection, and out to a photon detector. The radiator material used in the DIRC is 2.6. DETECTOR OF INTERNALLY REFLECTED LIGHT 35
PMT + Base 10,752 PMT's
Standoff Purified Water Light Catcher Box
17.25 mm Thickness (35.00 mm Width) Bar Box
Track PMT Surface Trajectory Wedge Mirror
Bar Window
4.9 m 1.17 m 4 x 1.225m Bars { glued end-to-end { 8-2000 8524A6
Figure 2.7: Diagram of DIRC. Distances are not to scale.
composed of bars of synthetic fused silica. The DIRC bars are arranged cylindrically around the drift chamber in a 12-fold symmetry.
The DIRC provides π/K separation greater than 4σ for momenta less than 3 GeV/c. Fig 2.8 shows the Cherenkov angle and timing resolution of the DIRC. Below 700 MeV/c particle ID is based primarily on the dE/dx measurements of the DCH and SVT. The particle ID is demonstrated using D0 Kπ events in Fig. 2.9. →
Additionally, the DIRC can be used to tag photons which shower before reaching the calorimeter. 36 CHAPTER 2. THE BABAR DETECTOR
80000 (a) 60000 40000 20000
entries per mrad 0 -100 -50 0 50 100 ∆ θ C,γ (mrad)
80000 (b) 60000 40000 20000
entries per 0.2ns 0 -5 0 5
∆ tγ (ns)
Figure 2.8: Cherenkov angle and timing resolution of the DIRC.
2.7 Electromagnetic Calorimeter
The electromagnetic calorimeter (EMC) is designed to detect and measure the en- ergies of photons and electrons of energies 20 MeV to 9 GeV with high precision and efficiency over a large angular region. The high light yield of CsI(Tl) crystals and high light collection efficiency allow for excellent photon energy resolution. The fine granularity of the detector allows accurate cluster position measurements which are crucial for π0 mass resolution at high π0 momentum, where it is dominated by angular
resolution. 2.7. ELECTROMAGNETIC CALORIMETER 37
Without DIRC x 10 2 1500 2
1000
entries per 5 MeV/c With DIRC 500
0 1.75 1.8 1.85 1.9 1.95 Kπ mass (GeV/c2)
Figure 2.9: Invariant mass plot of candidate D0 Kπ events before and after → applying loose kaon ID.
2.7.1 Design
The structure of the EMC is divided into two regions: a central angular region referred to as the barrel and an forward endcap. A backward endcap is not included in the design as most of the decay products from BB¯ decays are boosted in the forward direction. The EMC consists of 6580 trapezoidal CsI(Tl) crystals, 5760 of which are in the barrel. The length of the crystals varies from 16 to 17.5 radiation lengths
(X0,CsI = 1.85cm), where the the longer crystals are in the forward direction, where the maximum energy of a particle is higher. The front face dimensions are roughly
4.8 4.8cm2 and the longitudinal axis of the crystals point away from the beam spot × 38 CHAPTER 2. THE BABAR DETECTOR
to reduce leakage. A diagram of the EMC is provided in Fig. 2.10.
2359
1555 2295 External Support
1375 1127 1801 26.8˚ 920
38.2˚ 558 15.8˚ 22.7˚
Interaction Point 1-2001 1979 8572A03
Figure 2.10: Longitudinal section of the EMC. Lateral dimensions are shown in mm and angles in degrees.
The EMC electronics are designed to contribute negligibly to the overall energy resolution. This is achieved by using crystals with high light yield, highly efficient and uniform light collection and low-noise amplification. Crystals are wrapped in diffuse reflective material (TYVEK) and read out by a pair of 2 1 cm2 silicon PIN × diodes (Hamamatsu S2744-08).
2.7.2 Calibration and Performance
Frequent calibration of the EMC is necessary to reduce the contribution of the energy calibration to the energy resolution. Calibration of the electronics gain and removal of non-linearities is done by precision charge injection into the preamplifier input. Pedestal offsets are measured using random triggers in the absence of beam.
The calibration of signal crystal energies is done using two different techniques, 2.7. ELECTROMAGNETIC CALORIMETER 39
depending on the energy scale of interest. At low energies, crystals are calibrated using a radioactive source. A flourine-containing liquid is activated by a beam of neutrons
and then piped along the front faces of the crystal. The decay 19F + n 16 N + α, → 16N 16 O∗ + β, 16O∗ 16 O + γ occurs with a half-life of 7 seconds, leaving no → → residual radioactivity. At high energies, the single crystal calibration is performed with Bhabha events. The Bhabha calibration constants are determined by minimizing
the sum 2 k k ci E (θ, φ) χ2 = i i − dep (2.2) σk k P ! X k where the sum runs over the single crystal energies i of all clusters k, the cluster
k energy resolution is σi, and the deposited cluster energy Edep(θ, φ) is determined from MC. At intermediate energies the single crystal calibration constant is determined from a logarithmic interpolation between the constants determined by the radioactive source and Bhabha calibrations.
After the single crystal energy scale as been determined it is necessary to correct these energies of clusters for shower leakage. This correction is determined using π0 events and µµγ events at low energy and high energy, respectively. At this stage the calorimeter energy scale is correctly determined. A final empirical correction for pho- ton energies as a function of proximity to crystal edges, is applied. This correction is applied for photon energies greater than Eγ,LAB > 1 GeV, where calorimeter position resolution is sufficient to resolve sub-crystal features.
The energy resolution of the EMC, derived from photons from π0 decays at low energy, and µµγ events at high energy is
σ (2.30 0.03 0.3)% E = (1.35 0.08 0.2)%, (2.3) E 4 E(GeV) ⊕ p 40 CHAPTER 2. THE BABAR DETECTOR
where E and σE are the energy and RMS error of a photon, measured in GeV. This parameterization is derived from data such as is shown in Fig. 2.11.
π0 → γγ Bhabhas 0.06 χc → J/ψ γ MonteCarlo
E 0.04
/
E σ
0.02
0.02 10-1 1.0 10.0 3-2001 8583A41 Photon Energy (GeV)
Figure 2.11: Fractional energy resolution σE/E versus energy as derived from various sources, compared to Monte Carlo expectations.
The angular resolution of the BABAR calorimeter is determined largely by the
Moli`ere radius of CsI(Tl), which is the radius in which 90% of an electromagnetic shower is contained, and the transverse granularity of the calorimeter. The angular resolution is determined from symmetric π0 and η decays to be
4.16 0.04 σθ = σφ = 0.00 0.04 mrad. (2.4) E(GeV) ⊕ ! p A parameterization of the angular resolution as a function of photon energy is shown in Fig. 2.12. 2.8. SUPERCONDUCTING COIL 41
π0 → γγ 12 MonteCarlo ) d
a 8 r m (
θ σ 4
0 0 1 2 3 3-2001 8583A42 Photon Energy (GeV)
Figure 2.12: EMC angular resolution as a function of photon energy.
2.8 Superconducting Coil
Radially just outside of the EMC is the superconducting coil. To measure the mo- menta of charged particles, the silicon vertex tracker and the drift chamber are con- tained within a 1.5 T longitudinal magnetic field from a superconducting solenoid.
This magnetic field is provided using an operating current of 4596 A. The radial field does not exceed B = 0.15 T within the volume of the DCH. The photomultipliers | r| in the DIRC are sensitive to stray magnetic fields, requiring an additional magnet (the bucking coil) to offset these magnetic fields in the DIRC stand-off box, where
the DIRC photomultipliers are mounted. 42 CHAPTER 2. THE BABAR DETECTOR
2.9 Instrumented Flux Return
The instrumented flux return (IFR) is used to identify muons and neutral hadrons
0 (primarily KL and neutrons). A magnetic flux return is situated immediately outside the cryostat of the superconducting coil to prevent any radial component of the magnetic field in the tracking volume. This flux return is instrumented to provide muon identification. The detection of muons is important in bottomonium physics to measure decays Υ (nS) µ+µ−. As muons are minimum ionizing particles, they are → able to penetrate large amounts of material, including the EMC. The steel and brass of the flux return is used as a hadron absorber.
The IFR (shown in Fig. 2.13) consists of a barrel and two endcap sections. The barrel is composed of six sections, arranged in a hexagonal pattern, covering the central region of the detector. The forward regions is composed to four doors, each covering half of a forward region.
Barrel 3200 FW 920 3200
19 Layers
18 Layers BW 1940 1250