c 2006 by Paras P. Naik. All rights reserved.I must thank the many teachers from Addison schools who...
Transcript of c 2006 by Paras P. Naik. All rights reserved.I must thank the many teachers from Addison schools who...
c© 2006 by Paras P. Naik. All rights reserved.
MEASUREMENT OF THE RELATIVE AMPLITUDE AND STRONG PHASEBETWEEN D
0 → K∗+K− AND D0 → K∗+K−
VIA DALITZ PLOT ANALYSIS OF D0 → K+K−π0 DECAYS
BY
PARAS P. NAIK
B.S., University of Illinois at Urbana-Champaign, 2000M.S., University of Illinois at Urbana-Champaign, 2001
DISSERTATION
Submitted in partial fulfillment of the requirementsfor the degree of Doctor of Philosophy in Physics
in the Graduate College of theUniversity of Illinois at Urbana-Champaign, 2006
Urbana, Illinois
Abstract
I present physics concepts, which are useful to understand our analyses, and describe
the CLEO III and CLEO-c experiments at the Cornell Electron Storage Ring. I also
present motivations for a Dalitz plot analysis of the Cabibbo-suppressed charmed
meson decay mode D0 → K+K−π0 at CLEO. The analysis uses 9.0 fb−1 of data
collected at√
s ≈ 10.58 GeV with the CLEO III detector. We find the strong
phase difference δD ≡ arg(
A(D0→K∗+K−)
A(D0→K∗+K−)
)= 332◦ ± 8◦ ± 11◦ and relative amplitude
rD ≡∣∣∣A(D
0→K∗+K−)A(D0→K∗+K−)
∣∣∣ = 0.52 ± 0.05 ± 0.04. This measurement indicates significant
destructive interference between D0 → K+(K−π0)K∗− and D0 → K−(K+π0)K∗+
in the D0 → K+K−π0 Dalitz plot region where these two modes overlap. The fit
includes the K∗± and φ resonances and a non-resonant amplitude, and the measured
fit fractions for each resonance (with statistical uncertainty only) are (46.1 ± 3.1)%
for the K∗+, (12.3 ± 2.2)% for the K∗−, (14.9 ± 1.6)% for the φ, and (36.0 ± 3.7)%
for the non-resonant contribution. We find δD = 313◦ ± 9◦ (stat.) and an amplitude
ratio of rD = 0.52 ± 0.05 (stat.) from a second fit which substitutes scalar κ±
(mass 878 MeV/c2, width 499 MeV/c2) amplitudes for the non-resonant amplitude.
The measured fit fractions for each resonance (with statistical uncertainty only) are
(48.1 ± 4.5)% for the K∗+, (12.9 ± 2.6)% for the K∗−, (16.1 ± 1.9)% for the φ,
(12.6 ± 5.8)% for the κ+, and (11.1 ± 4.7)% for the κ−. We also investigate the
D0 → K+K−π0 Dalitz plot in 281 pb−1 of data collected at√
s ≈ 3.77 GeV with the
CLEO-c detector. We find results which are consistent with the CLEO III analysis.
I conclude by summarizing our results and present a brief appendix detailing the
K-matrix formalism.
iii
In memory of my grandfather,
Dr. Y.G. Naik (Physics)
Principal, Gujarat College
Dean, Faculty of Science
Gujarat University
Ahmedabad, Gujarat, India.
1906-1976
He would have loved to read this.
iv
Acknowledgments
I appreciate this opportunity to thank those people in my life who have brought out
the best in me, and thus directly helped me to pursue and complete the highest degree
awarded in Physics.
First, and foremost, I must thank my thesis advisor, Mats Selen. Mats has always
been available to discuss physics and assist me in my research, and he has always been
supportive of my personal and professional goals. Hopefully someday I will have the
opportunity to teach and guide someone towards success as well has he has.
I must thank my parents, Pradeep and Aruna, and my brother, Samar. I am
fortunate to have had the kind of support and friendship that they have provided me
over the years. I also thank my extended family and family friends who have always
treated me like their own son or brother.
I must thank Heather Gorman, without whom I wonder where I would have found
the motivation to complete my studies and move on to new challenges. Her patience,
affection, and consideration have helped me to achieve a lot in a short amount of
time. I am excited and pleased that the next phase of my life will begin in Ithaca,
where we first met.
I must thank the University of Illinois and the Department of Physics here, where I
have continued my education for the last ten years. I must also thank the University of
Illinois for awarding me a University Fellowship for the Fall 2005 semester, providing
me with the assistance to remain a research assistant when I returned from Ithaca.
I must thank the many collaborators I had while working on particle physics
research as part of the CLEO experiment. I must first thank the “original” (room) 487
v
crew of Scott Davis, Chris Sedlack, and Jeremy Williams for their helpful assistance
and endless supply of comedic one-liners. I must thank Norm Lowrey for helping
me acclimate myself to living and working in Ithaca. I must also thank Eric White,
who was pivotal in helping me assemble an additional chapter investigating the most
recent CLEO data.
I must thank Charles Plager for his help and support in understanding the impor-
tant physics and the relevant tools for studying Dalitz plots, and Tim Bergfeld whose
work has guided Illinois CLEO Dalitz plot analyses. I also must thank Bob Eisen-
stein, Jim Wiss, Inga Karliner, Doris Kim, and Topher Cawlfield for their assistance
with my research.
I must thank the entire Charm/Dalitz group for their help with various theoretical
and technical issues, including Steve Dytman who helped me to study the K-matrix
formalism. I would particularly like to thank Jim Napolitano, Jon Rosner, Mikhail
Dubrovin, and Hanna Mahlke-Kruger for all of their comments and assistance in
publishing a paper based on my research.
I must especially thank David Asner for his continuing assistance and involvement
with my research, and for offering me a job as a postdoctoral fellow with Carleton
University, so that I may continue to pursue scientific research in the field of elemen-
tary particle physics. I am looking forward to the opportunity.
I must thank the support staff for the Illinois High Energy Physics Group for
helping me arrange for travel and expenses related to research. I also must thank
the computing staffs at both Illinois and Cornell for providing me with the tools and
help necessary to complete computing projects.
I must thank my committee members, Kevin Pitts (chair), John Stack, and Brian
Fields, who have happily made their time and schedules available to remain apprised
of my research and encourage me to succeed.
I must thank the many teachers from Addison schools who have helped me along
vi
the way, especially: Tom Bookler, for helping me explore a world of curiosity from
elementary school through junior high; Walter Raczynski and especially Mark Ailes,
for piquing my interest in physics; and David Porter, for being a wonderful Calculus
teacher and writing a letter of recommendation that I am sure led to my first position
as a teaching assistant (TA).
I must thank the Department of Physics for giving me the opportunity to teach
as an undergraduate and as a graduate student. I appreciate the support of Gary
Gladding and Tim Stelzer during my years of involvement with the Physics Education
Research Group. I thank Johnetta Wilde and Cindy Hubert for their assistance and
conversations while I was a teaching assistant. I also thank all of my fellow TAs and
professors whom I worked with while teaching and as Mentor TA.
I must thank the friends and colleagues who have helped me and/or kept me
grounded throughout the years, including: Matthew Eisenbraun, Eric Rasmussen,
Melissa Headley, Nikhil Trivedi, Tolgar Alpagut, Lambros Tsonis, Eugene and Adele
Torigoe, Jim and Chimei Schneider, Andrew and Emily Meyertholen, Adam and
Laura Rasmussen, Joe Zimmerman, Amy Rund, John Phillips, Curtis Jastremsky,
Melissa Cole, Clinton Cronin, Matt Chasse, Pete and Adrienne Zweber, “Big” Mike
Bell, Ryan Small, Mike Watkins, Brian O’Shea, Evan Graves, Paco Jain, Kurt and
Jennifer Haberkamp, Steven Poulakos, Bob Drost, Brian Lang, Justin Hietala, Steve
Hugo, Jerry Codner and Patty McNally, Jocelyn and Nick Voss, Tim and Lynde Klein,
Dom Ricci, Selina Li, Mike Weinberger, Gregg and Jana Thayer, Tina Majethia, Matt
and Katie Shepherd, Christine Crane, Batbold Sanghi, Lisa Gay, Joscelyn Fisher, and
Ahren Sadoff.
I also thank anyone whom I have forgotten to mention and you, the reader of this
dissertation.
This material is based upon work supported by the Department of Energy under
Grant DE FG02 91ER40677.
vii
Table of Contents
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xx
1 Relevant Physics Concepts . . . . . . . . . . . . . . . . . . . . . . . 11.1 The Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Quark Decay and the CKM Matrix . . . . . . . . . . . . . . . . . . . 21.3 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.4 Violation of Discrete Symmetries . . . . . . . . . . . . . . . . . . . . 71.5 Unitarity and the CKM Angles . . . . . . . . . . . . . . . . . . . . . 81.6 Measuring γ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.7 Four-Momentum and Invariant Mass . . . . . . . . . . . . . . . . . . 131.8 Resonances and Interference . . . . . . . . . . . . . . . . . . . . . . . 141.9 The Dalitz Plot Formalism . . . . . . . . . . . . . . . . . . . . . . . . 14
2 Model for Three-body Decays . . . . . . . . . . . . . . . . . . . . . . 172.1 Breit-Wigner Formalism . . . . . . . . . . . . . . . . . . . . . . . . . 172.2 Decay Amplitudes for a Particular Spin . . . . . . . . . . . . . . . . . 202.3 Amplitudes on the Dalitz Plot . . . . . . . . . . . . . . . . . . . . . . 22
3 CLEO, CESR, and LEPP . . . . . . . . . . . . . . . . . . . . . . . . 243.1 LEPP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.2 CESR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.3 The CLEO III Detector . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 Silicon Vertex Detector . . . . . . . . . . . . . . . . . . . . . . 283.3.2 Drift Chamber . . . . . . . . . . . . . . . . . . . . . . . . . . 293.3.3 dE/dx Particle Identification . . . . . . . . . . . . . . . . . . 313.3.4 RICH Detector . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3.5 Crystal Calorimeter . . . . . . . . . . . . . . . . . . . . . . . . 373.3.6 Solenoid Coil . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.3.7 Muon Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . 383.3.8 Trigger and Data Acquisition Systems . . . . . . . . . . . . . 38
3.4 The CLEO-c Detector . . . . . . . . . . . . . . . . . . . . . . . . . . 40
viii
3.4.1 The ZD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423.5 Reconstructing Events . . . . . . . . . . . . . . . . . . . . . . . . . . 433.6 Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 D0 → K+K−π0 in CLEO III . . . . . . . . . . . . . . . . . . . . . . . 454.1 Current Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.2 Data Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474.3 Event Candidates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.4 The Dalitz Plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524.5 The Matrix Element . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.6 Method of Maximum Likelihood . . . . . . . . . . . . . . . . . . . . . 574.7 Efficiency and Background . . . . . . . . . . . . . . . . . . . . . . . . 60
4.7.1 Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.7.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.8 Overview of Fitting Technique . . . . . . . . . . . . . . . . . . . . . . 644.9 Goodness of Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.10 Fit Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.10.1 Nominal Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.10.2 κ Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744.10.3 Results of Nominal Fit and κ Fit . . . . . . . . . . . . . . . . 744.10.4 Other Fits to the Entire Dalitz Plot . . . . . . . . . . . . . . . 854.10.5 Fits to Partial Regions of the Dalitz Plot . . . . . . . . . . . . 994.10.6 Floating mκ, Γκ Fit to the Entire Dalitz Plot . . . . . . . . . . 113
4.11 Systematic Errors for rD and δD in our Nominal Fit . . . . . . . . . . 1164.12 Branching Ratio Cross-check . . . . . . . . . . . . . . . . . . . . . . . 1224.13 U -spin Symmetry Check . . . . . . . . . . . . . . . . . . . . . . . . . 1234.14 The Effect of our Uncertainties on γ . . . . . . . . . . . . . . . . . . 1244.15 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5 D0 → K+K−π0 in CLEO-c . . . . . . . . . . . . . . . . . . . . . . . . 1275.1 Effects of Quantum Correlations . . . . . . . . . . . . . . . . . . . . . 1275.2 Data Sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1305.3 Event Candidates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1315.4 CLEO-c Dalitz Plot Fit . . . . . . . . . . . . . . . . . . . . . . . . . 131
6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
A The K-matrix Formalism . . . . . . . . . . . . . . . . . . . . . . . . 138A.1 The K operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138A.2 The K-matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140A.3 K-matrix Examples and Argand Plots . . . . . . . . . . . . . . . . . 141
A.3.1 One Pole, One Decay Channel . . . . . . . . . . . . . . . . . . 141A.3.2 Two Poles, One Decay Channel . . . . . . . . . . . . . . . . . 142
A.4 Application to Dalitz Plot Analysis . . . . . . . . . . . . . . . . . . . 145A.4.1 Lorentz-Invariant T -Matrix . . . . . . . . . . . . . . . . . . . 145
ix
A.4.2 The �P -Vector Formalism . . . . . . . . . . . . . . . . . . . . . 146A.4.3 Practical Application . . . . . . . . . . . . . . . . . . . . . . . 147
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Author’s Biography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
x
List of Tables
2.1 Normalized Blatt-Weisskopf Barrier Factors (FP (mab) = 1 when mab =mr) for the particle P decay vertex. zP is defined as follows: zP (mab) =r2P p2
P (mab). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1 Branching ratios of intermediate D0 modes which may decay to thethree-body mode K+K−π0. . . . . . . . . . . . . . . . . . . . . . . . 46
4.2 The masses and widths of resonances r considered in this analysis[1, 45–48]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3 Best-fit parameters for the efficiency shape. . . . . . . . . . . . . . . . 614.4 Best-fit parameters for the background shape. . . . . . . . . . . . . . 644.5 Summary of fits. See Table 4.2 for more information about each reso-
nance, and see the following figures for details. . . . . . . . . . . . . . 714.6 Amplitude, phase, and fit fraction results for our nominal fit (Fit 1)
with statistical errors. . . . . . . . . . . . . . . . . . . . . . . . . . . 764.7 Amplitude, phase, and fit fraction results for Fit 3 (with κ± instead of
a non-resonant contribution) with statistical errors. . . . . . . . . . . 854.8 Summary of systematic checks and systematic errors. “<” means a
small change. * RICH selection criteria are used when RICH info isavailable for kaons with momentum pK > 500 MeV. ** dE/dx selec-tion criteria are used when pK < 500 MeV or when RICH info is notavailable for kaons with momentum pK > 500 MeV. . . . . . . . . . 118
4.9 Details about systematic checks. * RICH selection criteria are usedwhen RICH info is available for kaons with momentum pK > 500 MeV.** dE/dx selection criteria are used when pK < 500 MeV or whenRICH info is not available for kaons with momentum pK > 500 MeV. 120
4.10 Best-fit parameters for the background shape when the backgroundparameters are allowed to float. . . . . . . . . . . . . . . . . . . . . . 121
4.11 Best-fit parameters for the efficiency shape when the efficiency param-eters are allowed to float. . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.12 Comparison of estimated branching ratios from this analysis to pub-lished branching ratios [32–34,48] based on our nominal Dalitz plot fit.We accounted for the fact that K∗± decays to K±π0 only one-third ofthe time, and φ decays to K+K− only about one-half of the time. . . 123
5.1 Amplitude, phase, and fit fraction results for Fit 1-c with statisticalerrors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
xi
5.2 Amplitude, phase, and fit fraction results for Fit 3-c with statisticalerrors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.3 Comparison of predicted to measured ratios of fit fractions when com-paring CLEO-c data (correlated D0s) to CLEO III data (uncorrelatedD0s). The errors on the predictions are propagated from the totalerrors on rK∗K and δK∗K . The errors on the ratios from data are sta-tistical errors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
A.1 Partial widths used to calculate �F for interfering K∗ resonances. . . . 150
xii
List of Figures
1.1 The quarks and leptons of the Standard Model. Each row consists ofparticles with the same electric charge and the columns are groupedinto three generations each. . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 One example of a D0 → K+K−π0 decay. . . . . . . . . . . . . . . . . 41.3 The electromagnetic interaction e+e− → qq. . . . . . . . . . . . . . . 51.4 The Unitarity Triangle showing the common definitions of the three
weak phases α (also referred to as φ2), β (φ1), and γ (φ3), as well asthe relevant CKM matrix elements. . . . . . . . . . . . . . . . . . . . 9
2.1 D0 → abc through an ab resonance. The initial and final statesboth have no angular momentum, so the intermediate spin states aresummed over. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 Since the D0 is a spinless particle, a vector (spin 1) decay particlesuch as the φ must be in a |1, 0〉 angular momentum state in the D0
rest frame (when we quantize along the π0 direction). The daughtersof the φ therefore decay preferentially parallel to its momentum (theamplitude has a Y 1
0 ∝ cos(θ) dependence in the D0 rest frame). . . . 222.3 |A|2 for the vector K−π0 resonance K∗−. When the π0 comes out
opposite to the K∗− direction, it is almost at rest and corresponds toθ = 0◦. The K− is almost at rest when θ = 180◦. When the K−
and π0 come out perpendicular to the K∗− (θ = 90◦), they have equalmomentum. We can see the preference for a decay to occur parallel tothe momentum of the K∗− as a cos2(θ) dependence on the Dalitz plot(the square of the cos(θ) amplitude dependence mentioned in Figure2.2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 A diagram of CESR. . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2 The CLEO III Detector. . . . . . . . . . . . . . . . . . . . . . . . . . 273.3 End view and quarter-section view of the CLEO III Silicon Detector. 283.4 Charge collection and multiplication in the drift chamber. . . . . . . . 293.5 Using tilted stereo wires to obtain z information in the drift chamber.
Tilted wires are projected as arrows coming out of the page in the r-φprojection; position along a tilted wire indicates the z of the track helixnear that wire. The closest wires to the track (in three dimensions)are highlighted (in green). . . . . . . . . . . . . . . . . . . . . . . . . 30
xiii
3.6 CLEO III dE/dx vs. momentum showing the π, K and p bands.Leptons are suppressed where possible. The dE/dx scale is chosen tobe 1 for electrons and positrons from e+e− → e+e− events. . . . . . . 32
3.7 A section of the RICH detector. . . . . . . . . . . . . . . . . . . . . . 343.8 How true kaons (simulated) would look at various momenta (binned)
in terms of the range of possible values for Δχ2dE/dx (Equation 3.2)
for dE/dx information and the range of possible values for Δχ2RICH
(Equation 3.5) for RICH information, as defined in the text. Values ofless then zero indicate that the kaon hypothesis is more likely. . . . . 36
3.9 How the trigger tiles read shower information from the calorimeter. . 393.10 Cross section of the CLEO-c detector in the r-z plane. . . . . . . . . . 413.11 An isometric view of the CLEO-c inner wire chamber. . . . . . . . . . 42
4.1 Shown are examples of (a) a non-resonant decay, (b) a neutral inter-mediate state (φ), (c) a positively charged intermediate state (K∗+),and (d) a negatively charged intermediate state (K∗−). . . . . . . . . 46
4.2 Histograms of (−2 lnLK)−(−2 lnLπ) from RICH information, numberof photons nγ used to determine the RICH information, and σK , thenumber of standard deviations away from the expected dE/dx for atrue kaon, for the positive (1st row) and negative (2nd row) kaons afterall other selection criteria have been applied. . . . . . . . . . . . . . . 49
4.3 Distribution of (a) mK+K−π0 for |ΔM | < 1 MeV/c2 and (b) ΔM for1.84 < mK+K−π0 < 1.89 GeV/c2 after passing all other selection cri-teria discussed in the text. The solid curves (red) show the results offits to the mK+K−π0 and ΔM distributions, respectively. The back-ground level in each plot is shown by a dashed horizontal line (black).The vertical lines in (a) and the left-most set of vertical lines in (b)denote the signal region. The right-most set of vertical lines in figure(b) denote the ΔM sideband used for estimation of the backgroundshape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.4 (a) The Dalitz plot distribution for D0 → K+K−π0 candidates. (b)-(d) Projections onto the m2
K+π0 , m2K−π0 , and m2
K+K− axes of the resultsof Fit 1 (discussed in Section 4.10) showing both the fit (curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 53
xiv
4.5 An example of the effect of a non-resonant (NR) contribution interfer-ing with two K∗ resonances. The K∗s in this example have the sameamplitude but a relative phase with each other such that the two reso-nances have maximal destructive interference with each other, and theamplitude aNR is set to be 5aK∗ . When the relative phase of the NRcontribution to the K∗+ phase, δNR, is 0◦ or 180◦ there is not a notice-able effect on the K∗ lobes. However, a phase δNR = 90◦ makes theK∗− enhanced (depleted) at low (high) K+π0 invariant mass squared,and the K∗+ enhanced (depleted) at high (low) K−π0 invariant masssquared. A phase δNR = 270◦ makes the K∗+ enhanced (depleted)at low (high) K−π0 invariant mass squared, and the K∗− enhanced(depleted) at high (low) K+π0 invariant mass squared. . . . . . . . . 54
4.6 Dalitz plot projections properly rotated to show how they are obtainedfrom the Dalitz plot. The thicker straight lines correspond to reso-nances which decay to the two particles in the respective invariantmass-squared projection, and the thinner straight lines correspond toreflections which do not decay to the two particles in the respectiveinvariant mass-squared projection. . . . . . . . . . . . . . . . . . . . . 56
4.7 (a) Scatter plot of Monte Carlo events used to study the efficiencyacross the Dalitz plot and (b-d) projections of these events along withthe result of a fit using the efficiency function. . . . . . . . . . . . . . 62
4.8 (a) Scatter plot of ΔM versus mK+K−π0 for event candidates satisfyingall other requirements. The black boxed area (the bottom box) is theSignal Region which contains our signal candidates, and the red boxedarea (the top box) is the Background Region which contains eventswhich we use to estimate the shape of our background. (b) a Dalitzplot of the events from the Background Region of plot (a). . . . . . . 63
4.9 The projected background events from the ΔM sideband and the resultof fitting the background Dalitz plot to a 2-D cubic polynomial plusnon-interfering K∗ and φ contributions (line). . . . . . . . . . . . . . 65
4.10 Interfering K∗+ and K∗− resonances on the Dalitz plot. The waywe have defined our amplitudes results in a phase convention wheremaximal destructive interference between the K∗s occurs at δD = 0◦.Maximal constructive interference occurs when δD = 180◦. Note thatthe effect of the interference is most noticeable in the region where theK∗ resonances overlap. . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.11 Dalitz plot projections for Fit 1 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.12 Dalitz plot projections for Fit 2 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.13 Dalitz plot projections for Fit 3 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.14 The signal and background in the Dalitz plot projections of Fit 1. . . 77
xv
4.15 Signal portion of Fit 1 and the component resonance amplitudes squared(m2
K+π0 projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784.16 Signal portion of Fit 1 and the component resonance amplitudes squared
(m2K+K− projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.17 Signal portion of Fit 1 and the component resonance amplitudes squared(m2
K−π0 projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.18 The signal and background in the Dalitz plot projections of Fit 3. . . 814.19 Signal portion of Fit 3 and the component resonance amplitudes squared
(m2K+π0 projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.20 Signal portion of Fit 3 and the component resonance amplitudes squared(m2
K+K− projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.21 Signal portion of Fit 3 and the component resonance amplitudes squared
(m2K−π0 projection). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.22 Fit 1 projections (blue, dashed line) overlayed on top of Fit 3 projec-tions (green, solid line). . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.23 Dalitz plot projections for Fit 3a showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.24 Dalitz plot projections for Fit 4 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.25 Dalitz plot projections for Fit 5 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.26 Dalitz plot projections for Fit 6 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.27 Dalitz plot projections for Fit 7 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.28 Dalitz plot projections for Fit 8 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.29 Dalitz plot projections for Fit 9 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.30 Dalitz plot projections for Fit 10 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.31 Dalitz plot projections for Fit 11 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.32 Dalitz plot projections for Fit 12 showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.33 Fit Regions (a) K∗ bands (b) No φ 1.05 (c) No φ 1.15 (d) No φ 1.26(e) “L” area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.34 Dalitz plot projections for Fit 1x showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.35 Dalitz plot projections for Fit 2x showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.36 Dalitz plot projections for Fit 3x showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 104
xvi
4.37 (a) The absolute value of the phase distribution of our nominal fit overthe entire Dalitz plot subtracted from the phase distribution of thenominal fit with the φ simply removed (aφ is set to be 0). This showsthe effect the φ resonance has on the overall phase at different pointsin phase space on the Dalitz plot. Removing only a region where theφ has its peak is not enough to remove all regions of the Dalitz plotwhose phase is noticeably affected by the φ resonance. (b) The sameplot as (a) except that areas outside the K∗ bands are excluded. Notethat the phase distribution is not affected by the φ in the vast majorityof the K∗ bands region, so it is definitely appropriate to exclude the φfrom any K∗ bands fit. . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.38 Dalitz plot projections for Fit 1n showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.39 Dalitz plot projections for Fit 1np showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.40 Dalitz plot projections for Fit 1na showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.41 Dalitz plot projections for Fit 1nap showing both the fit (red curve)and the binned data sample. . . . . . . . . . . . . . . . . . . . . . . . 110
4.42 Dalitz plot projections for Fit 1nb showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.43 Dalitz plot projections for Fit 1nbp showing both the fit (red curve)and the binned data sample. . . . . . . . . . . . . . . . . . . . . . . . 112
4.44 Dalitz plot projections for Fit 1s showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.45 Dalitz plot projections for Fit 3f showing both the fit (red curve) andthe binned data sample. . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.1 Plots of mbc and ΔE for all events passing the selection criteria inSection 5.2, with the signal region within the dashed (brown) lines ofboth figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.2 Dalitz plot of K+K−π0 candidates. . . . . . . . . . . . . . . . . . . . 1335.3 Dalitz plot projections for Fit 1-c. . . . . . . . . . . . . . . . . . . . . 1345.4 Dalitz plot projections for Fit 3-c. . . . . . . . . . . . . . . . . . . . . 135
A.1 (left) An Argand Plot of the K∗(892) T -matrix for all values of m2Kπ.
(right) A plot of |T |2 vs. m2Kπ. . . . . . . . . . . . . . . . . . . . . . . 142
A.2 (left) An Argand Plot of the K∗(1410) T -matrix for all values of m2Kπ.
(right) A plot of |T |2 vs. m2Kπ. . . . . . . . . . . . . . . . . . . . . . . 143
A.3 (left) An Argand Plot of T = TK∗(892) + TK∗(1410) for all values ofm2
Kπ < 5. (right) A plot of |TK∗(892) + TK∗(1410)|2 vs. m2Kπ. . . . . . . 143
A.4 (left) An Argand Plot of T calculated from KK∗(892) + KK∗(1410) for allvalues of m2
Kπ < 5. (right) A plot of |T |2 vs. m2Kπ. . . . . . . . . . . . 144
xvii
A.5 A diagram of the process where a resonance is produced from a D0
decay. The resonance may decay to a number of decay channels, ifpossible. The amplitude for the decay channel with particles a and bcan be found in the ab component of the final state vector �F . . . . . . 146
A.6 A plot of the magnitude of the �F -vector amplitude to the Kπ decaychannel squared vs. Kπ invariant mass squared for destructively in-terfering K∗ resonances. . . . . . . . . . . . . . . . . . . . . . . . . . 150
A.7 An example of a fit using the P -vector formalism. . . . . . . . . . . . 152
xviii
List of Abbreviations
APS American Physical Society.
BR Branching Ratio.
C Charge Conjugation.
CESR Cornell Electron Storage Ring.
CHESS Cornell High Energy Synchrotron Source.
CKM Cabibbo-Kobayashi-Maskawa.
CP Charge-Parity.
DAQ Data Acquisition System.
DP Dalitz Plot.
FF Fit Fraction.
LEPP Laboratory for Elementary-Particle Physics.
LINAC Linear Accelerator.
MC Monte Carlo.
NR Non-resonant.
P Parity Inversion.
Ph.D. Doctor of Philosophy.
RF Radio Frequency.
RICH Ring Imaging Cherenkov.
S.L. Significance Level.
Stat. Statistical Error.
Std. Dev. Standard Deviations.
T Time Reversal.
Vs. Versus.
xix
List of Symbols
G Giga. 109
M Mega. 106
k Kilo. 103
c Centi. 10−2
m Milli. 10−3
μ Micro. 10−6
n Nano. 10−9
p Pico. 10−12
f Femto. 10−15
B Bytes. (A unit representing data)
b Barns. (A unit representing area, 1 b = 10−28 m2)
m Meters. (A unit representing length)
eV Electron Volts. (A unit representing energy, 1 eV ≈ 1.6 × 10−19 Joules)
Hz Hertz. (A unit representing frequency, 1 Hz = 1 s−1)
s Seconds. (A unit representing time)
◦ Degrees. (A unit representing angle)
rad Radians. (A unit representing angle, 2π rad = 360◦)
O Order
→ Decays To
�x Vector x
x Anti-particle of x
X Operator X
X Lorentz Invariant Version of X
xx
1 Relevant Physics Concepts
1.1 The Standard Model
The Standard Model of particle physics, which summarizes our current understanding
of matter, includes six quarks and six leptons grouped into three generations (see
Figure 1.1) [1]. All of the quarks and leptons are spin 12
fermions. Each of the
particles in Figure 1.1 has an antiparticle which has the same mass and spin but a
change in sign of additive quantum numbers, most importantly electric charge (d has
charge − e3, d has charge + e
3; −e is the electric charge of the electron). The quarks and
leptons, along with their antiparticles, are assumed to be the fundamental particles
of matter in the theory.
There are three forces that affect the interaction of particles: strong, electroweak,
and gravity. The effects of gravity are neglected in the Standard Model because of
its relative weakness in comparison to the other forces [2]. The electroweak force at
current experimental energies can be thought of in terms of separate electromagnetic
and weak interactions. Quarks interact via the electromagnetic, weak, and strong
chargecharge
+2e3− e3
(ud
) (cs
) (tb
)6 Quarks, 3 Generations
chargecharge
0−e
(νe
e
) (νμ
μ
) (ντ
τ
)6 Leptons, 3 Generations
Figure 1.1: The quarks and leptons of the Standard Model. Each row consists ofparticles with the same electric charge and the columns are grouped into three gen-erations each.
1
forces; charged leptons only interact through the electromagnetic and weak forces, and
neutral leptons only interact via the weak force [1]. The electromagnetic interaction
is mediated by a massless, spin-1 photon. The weak interaction is mediated by the
massive, spin-1, charged W± and neutral Z bosons. The strong force is mediated by
massless, spin-1 gluons.
Isolated leptons can readily be found in nature, but quarks have only been ob-
served in bound states of two types: mesons, which consist of a quark bound to an
antiquark (qq); and baryons, which consist of three quarks (qqq) or three antiquarks
(qqq). The mesons and baryons are collectively known as hadrons due to their ability
to interact via the strong force [2].
1.2 Quark Decay and the CKM Matrix
The weak interaction allows one type of quark to change into another type of quark,
also known as changing flavor, by emitting or absorbing a charged W boson. Gener-
ally, a quark will decay within its generation unless forbidden by energy conservation.
Less frequently, a particle will decay into another generation. This ability to decay
into another generation is due to a difference in the weak and mass eigenstates for the
quarks. A charge − e3
quark’s weak eigenstates (|d′〉, |s′〉, |b′〉) are related to its mass
eigenstates (|d〉, |s〉, |b〉) by the Cabibbo-Kobayashi-Maskawa (CKM) quark-mixing
matrix, VCKM in Equation 1.1 [1]. We have the freedom to define the charge +2e3
quarks’ weak eigenstates to be the same as their mass eigenstates (Equation 1.2), so
that the VCKM contains all of the information about the quark mixing.
⎛⎜⎜⎜⎜⎝
d′
s′
b′
⎞⎟⎟⎟⎟⎠ = VCKM
⎛⎜⎜⎜⎜⎝
d
s
b
⎞⎟⎟⎟⎟⎠ =
⎛⎜⎜⎜⎜⎝
Vud Vus Vub
Vcd Vcs Vcb
Vtd Vts Vtb
⎞⎟⎟⎟⎟⎠
⎛⎜⎜⎜⎜⎝
d
s
b
⎞⎟⎟⎟⎟⎠ (1.1)
2
⎛⎜⎜⎜⎜⎝
u′
c′
t′
⎞⎟⎟⎟⎟⎠ =
⎛⎜⎜⎜⎜⎝
u
c
t
⎞⎟⎟⎟⎟⎠ (1.2)
The coupling strength used in determining a quark’s interaction probability am-
plitude is defined by gVij where i and j are types (or “flavors”) of quarks and g is the
universal weak coupling constant [3]. An antiquark’s interaction probability ampli-
tude is the complex conjugate, gV ∗ij . The diagonal elements of the CKM matrix are
close to 1 since quarks prefer to decay to their own generation if allowed; however,
all of the off-diagonal elements have magnitudes that are non-zero within 90% con-
fidence limits. Conservation of probability requires the CKM matrix to be unitary,
assuming there are only 3 generations of quarks. The current global fit which uses
all available measurements of Vij and the unitarity constraint [1] gives the following
for the magnitudes of the CKM elements:
VCKM =
⎛⎜⎜⎜⎜⎝
0.97383+0.00024−0.00023 0.2272+0.0010
−0.0010 (3.96+0.09−0.09) × 10−3
0.2271+0.0010−0.0010 0.97296+0.00024
−0.00024 (42.21+0.10−0.80) × 10−3
(8.14+0.32−0.64) × 10−3 (41.61+0.12
−0.78) × 10−3 0.999100+0.000034−0.000004
⎞⎟⎟⎟⎟⎠ . (1.3)
It should be noted that elements of the CKM matrix are allowed to be complex
numbers.
Two examples of how the weak interaction occurs can be seen in a decay we are
interested in, D0(cu) → K+(us)K−(us)π0(uu−dd√2
). The particular Feynman diagram
shown in Figure 1.2 is just one of many that could be drawn for the D0 → K+K−π0
decay. The first weak decay, c → W+s, allows the c quark to change into an s quark.
The second weak decay occurs when the W+ boson decays into a u and s quark.
Decays that involve quarks within the same generation, such as c → W+s are known
3
c
u_
s
u_
u_
D0
W+
K-
K+
π0
u
u
s_
gg
Figure 1.2: One example of a D0 → K+K−π0 decay.
as Cabibbo-favored decays since the diagonal CKM matrix elements are relatively
large, while decays that involve quarks of different generations, such as W+ → us,
are known as Cabibbo-suppressed. It turns out that in all possible Feynman diagrams
for the D0 → K+K−π0 decay, there must be at least one Cabibbo-suppressed decay,
and thus the decay itself is referred to as Cabibbo-suppressed also.
Mesons may also decay by way of the strong force. Often a meson may be in
a highly energetic state, such as when individual quarks of the meson are traveling
quickly enough to separate. It is energetically preferred to create new quark-antiquark
pairs in order to lower the energy of the original meson by forming two separate
particles [4]. An example of this type of decay in Figure 1.2 occurs after the W+ → us
vertex, when a uu pair is created to hadronize the separating u and s quarks.
The strong force is itself responsible for the existence of hadrons. Quarks have a
property called “color,” which is conserved in strong interactions (this is an analog
to charge which is conserved in electromagnetic interactions). Color is a necessary
4
e+
e-
q
q_
γ∗
Figure 1.3: The electromagnetic interaction e+e− → qq.
quality of quarks, because, without it, protons (uud) would not be able to have two
u quarks in the same state due to the Pauli exclusion principle for fermions. Quarks
may have three types of colors, antiquarks have three types of anticolors, and gluons
carry color-anticolor combinations. Bound states of quarks (hadrons) must have no
net color. This is the case when quarks of all three colors (or all three anticolors)
form a bound state (baryons) or when a quark has one color and an antiquark has
the anticolor (mesons).
A relevant lepton electromagnetic interaction is that of an electron (e−) − positron
(e+) annihilation into a virtual photon, which then produces a quark-antiquark pair
(see Figure 1.3). This interaction is utilized in some particle accelerators (such as
the Cornell Electron Storage Ring) to create qq pairs which separate, hadronize, and
decay to create mesons and baryons that particle physicists are interested in.
1.3 Symmetries
In physics, there are certain fundamental symmetries which lead to conservation (in-
variance) of physical quantities. Some of these are continuous symmetries, while
others are non-continuous, or discrete. An example of a continuous symmetry is time
translation, where a physical system at time t is compared to an identical physical
system at time t + Δt, where Δt is an non-zero interval of time. If we are stand-
5
ing at sea level and throw a ball up vertically (neglecting air resistance) at time t,
Newton’s Second Law, �F = m�r, would correctly predict the ball’s acceleration under
the influence of gravity. If we throw the ball at time t + Δt under the same ini-
tial conditions, we still expect Newton’s Second Law to provide the same result for
the acceleration. Newton’s Second Law is unchanged under a time translation; the
acceleration is preserved under this symmetry.
An example of a discrete symmetry is time reversal (T ), where t is replaced with
−t. This replacement would reverse velocities, momenta, and angular momenta, but
would not reverse displacements or accelerations. Newton’s Second Law still holds
under this transformation. Imagine the motion of the ball in the previous example
thrown upward at time t. If we reversed time by applying the T symmetry, we would
see that the reverse process would appear to be the same, the process being symmetric
with respect to the moment at which the ball is at it’s maximum height.
Another discrete symmetry is parity inversion (P ), which is the spatial inversion
of position with respect to the origin. We can define a parity operator, P , which
performs this operation:
P�r = P (x, y, z) = (−x,−y,−z) = −�r. (1.4)
Velocities, accelerations, and momenta will all be reversed under parity inversion, but
angular momenta will not. Angular momentum is a vector product of a displacement
and a momentum, and is preserved under a parity inversion:
P �L = P (�r × �p) = −�r ×−�p = �r × �p = �L. (1.5)
Quantities such as �r which change sign under parity transformation have a parity
eigenvalue, P = −1. Quantities such as �L which do not change sign under parity
6
transformation have a parity eigenvalue, P = 1, and are called Pseudovectors. Note
that applying the parity operator twice is the same as applying the identity operator,
P 2 = I. Hadrons are eigenstates of the parity operator [5]. The convention is that
quarks have positive parity (+1), and anti-quarks have negative parity (−1). The
parity of a composite system of quarks is the product of the parities of each quark
multiplied by a factor of (−1)L, where L is the orbital angular momentum of the
hadron. Also, the photon is a vector (spin 1) particle and has an intrinsic parity
P = −1.
One more example of a discrete symmetry is charge conjugation symmetry (C).
The charge conjugation operator, C, reverses all internal quantum numbers, including
electric charge, thus turning a particle into its anti-particle. For example, C|K+〉 =
|K−〉. Most hadrons are not eigenstates of C, but those that are (|π0〉, for example)
have charge conjugation quantum number C = (−1)L+S, where S is the spin of the
hadron. Like P , applying C twice to a particle returns the same particle, thus C2 = I.
1.4 Violation of Discrete Symmetries
C and P symmetries hold under strong decays, but are violated in weak decays [1].
Charge-Parity (CP ) symmetry is the combination of C symmetry and P symmetry.
CP symmetry was once thought to hold true for weak decays, but a violation of this
symmetry has been observed in neutral kaons and the neutral B-meson system [1].
The term “CP violation” refers to violation of the combined CP symmetry. CP
violation has not yet been observed in decays of D-mesons. In the case that CP
violation is negligible, the rates of conjugate decays are equal (for example, we would
expect as many D0 mesons to decay to K∗−K+ as we would expect D0(cu) mesons
to decay to K∗+K−). This is an acceptable assumption for this analysis, because CP
violation is predicted to be negligibly small for D decays in the Standard Model [1].
7
One of the main motivations for a better understanding of, and the continued
search for, CP violation is to explain why matter is preferred over anti-matter in
our universe. While CP violation is not sufficient, it is necessary to explain this
dominance of matter [6]. There are few tests of processes in which CP symmetry
is violated. When there is a violation, such as for neutral kaons, it is consistent
with the Standard Model. However, the Standard Model violation is not a large
enough asymmetry to explain a matter-dominated universe. Looking deeper into CP
violation and its parameters may lead to observations of physics beyond the Standard
Model which would explain the matter-antimatter asymmetry.
So far, no experimental evidence for CPT violation, the violation of the combined
C, P , and T symmetries, has been observed [1].
1.5 Unitarity and the CKM Angles
We have the freedom to parameterize the CKM matrix in terms of a single phys-
ical phase because it is possible to redefine the quark mass eigenstates to absorb
other phases. An explicit parametrization of this type is known as the Wolfenstein
parametrization [1]:
VCKM =
⎛⎜⎜⎜⎜⎝
1 − λ2
2λ Aλ3(ρ − iη)
−λ 1 − λ2
2Aλ2
Aλ3(1 − ρ − iη) −Aλ2 1
⎞⎟⎟⎟⎟⎠ + O(λ4), (1.6)
where higher order terms in λ are ignored because λ is small and real. A, ρ, and η
are real numbers of order unity.
We expect that the inner product of any two of the columns, or any two of the
rows, of the CKM matrix is zero because a unitary matrix is orthogonal. We may
apply this unitarity constraint to the first and third columns of the CKM matrix [1]
8
γ
α
β
*
Vud
Vub
* *Vtd
Vtb
Vcd
Vcb
Figure 1.4: The Unitarity Triangle showing the common definitions of the three weakphases α (also referred to as φ2), β (φ1), and γ (φ3), as well as the relevant CKMmatrix elements.
which yields:
VudV∗ub + VcdV
∗cb + VtdV
∗tb = 0. (1.7)
Equation 1.7 is interesting because Vub and Vtd are both matrix elements that are
complex in the O(λ2) Wolfenstein parametrization. A unitarity triangle (Figure 1.4)
can represent Equation 1.7 in the complex plane [1].
The size of the area, JCP
2≈ λ6A2η
2, of the triangle in Figure 1.4 determines the
amount of charge-parity symmetry (CP ) violation in weak quark decays and in the
current global fit is measured to be JCP
2= (1.54+0.08
−0.09) × 10−5 [1]. There are five
additional triangles that can be made which enable us to overconstrain the area if we
have information about the angles and sides of the triangles. The CKM angles α (also
referred to as φ2), β (φ1), and γ (φ3) of Figure 1.4 can, in principle, be measured by
CP asymmetries in B-meson decays, while the sides can also be measured by decays
involving the associated quarks.
9
1.6 Measuring γ
γ is a challenging CKM angle to determine experimentally. Of course, we can con-
strain γ with the knowledge of α and β assuming that the unitarity triangle is closed,
but a direct measurement would provide a test of unitarity itself and thus would be
sensitive to physics beyond the Standard Model.
There are a few promising ways in which γ may be measured directly from B-
meson decays. Some of these methods include measuring decays whose amplitudes
have only a relative CKM phase γ and a relative strong phase δ, measuring B →ππ, Kπ, or KK decays, partial reconstruction of B → D(∗)±π∓ decays, and measuring
Bs decays such as Bs → ρ0K0S and Bs → D0φ [7].
A method of measuring γ via B±(ub, ub) → DK± decays where the neutral D-
meson decays to CP eigenstates was proposed by Gronau and Wyler [8]. Atwood,
Dunietz, and Soni show that choosing final states which are not CP eigenstates
can lead to large direct CP violation which can also give significant information
about γ [9]. Grossman, Ligeti, and Soffer [10] have proposed a method for a direct
measurement of γ by studying B± → DK±, where the neutral D meson (D0/D0)
decays to the Cabibbo-suppressed final states K∗+K− or K∗−K+. The final states
both have sizable rates in D decay and are not CP eigenstates.
In this formalism (where we ignore terms of O(λ4) and higher order terms in λ),
we define:
A∓B ≡ A(B∓ → D0K∓) (1.8)
A∓B ≡ A(B∓ → D0K∓) (1.9)
A∓B
A∓B
= rBei(δB∓γ), (1.10)
10
and
A∓D ≡ A(D0 → K∗±K∓) (1.11)
A∓D ≡ A(D
0 → K∗±K∓) (1.12)
A∓D
A∓D
= rDeiδD , (1.13)
where A is the decay amplitude for the respective decay, δB is the relative strong
phase between A+B and A+
B, the CKM angle γ is the weak phase between A+B and
A+B, δD is the relative strong phase between A−
D and A−D, and rB and rD are real
and positive. There is no weak phase between A−D and A−
D since D0 → K∗+K− and
D0 → K∗+K− both have amplitudes proportional to V ∗csVus(= VcsV
∗us, since Vus and
Vcs are real up to O(λ3)).
Four interesting B-decay amplitudes can then be defined as follows:
A[B− → K−(K∗+K−)D] = |ABAD|(1 + rBrDei(δB+δD−γ)) (1.14)
A[B− → K−(K∗−K+)D] = |ABAD|eiδD(rD + rBei(δB−δD−γ)) (1.15)
A[B+ → K+(K∗−K+)D] = |ABAD|(1 + rBrDei(δB+δD+γ)) (1.16)
A[B+ → K+(K∗+K−)D] = |ABAD|eiδD(rD + rBei(δB−δD+γ)), (1.17)
where we define AB ≡ A−B and AD ≡ A−
D. The CKM angle γ can then be determined
with the knowledge of all of the other, measurable quantities in Equations 1.14, 1.15,
1.16, and 1.17.
An important ingredient in this proposition is the knowledge of the relative com-
plex amplitude rDeiδD between AD and AD, which, in the absence of CP viola-
tion in D decays, is the same as the relative complex amplitude between A+D =
11
A(D0 → K∗−K+) and AD = A(D0 → K∗+K−). The primary objective of the anal-
ysis described in this dissertation is to measure rD and δD via the three-body decay
D0 → K+K−π0. We expect the amplitude for K∗+ → K+π0 to be the same as the
amplitude for K∗− → K−π0. In this case, rDeiδD can be determined by the interfer-
ence between the D0 → K+(K−π0)K∗− and D0 → (K+π0)K∗+K− decay modes, as
discussed in Chapter 4.
Similar analyses can also be done with B− → K−(ρ±π∓)D decays [10], B → DK∗0
decays [7], and B− → K−Dmulti−body (where the D-meson decays to a multi-body final
state which both the D0 and D0
have large probability to decay to, such as K0Sπ+π−)
[11]. The latter method has already been used to make a direct measurement of
γ, with the Belle Collaboration measuring γ = (68+14−15 ± 13 ± 11)◦ and the BaBar
Collaboration measuring γ = (67 ± 28 ± 13 ± 11)◦, where the last uncertainty is
due to the D decay modeling [1]. A more recent result by Belle using multiple
B+ → D(∗)K(∗)+ decay modes is γ = (53+15−18 ± 3 ± 9)◦ [12].
Determining the length of sides of the unitarity triangle can also be a helpful way
to constrain γ, if our assumptions are correct and the unitarity triangle is closed. The
sides of the unitarity triangle corresponding to VudV∗ub and VcdV
∗cb can be measured
in b quark decays to u or c quarks, respectively, such as in b → u(c)lν, where l is
a charged lepton and ν is the corresponding neutrino [1]. The side of the unitarity
triangle corresponding to VtdV∗tb can be measured via measurements of B0-B
0and
B0s -B
0
s oscillations [1].
As of October 2006, the CKM angle γ is inferred to be 59.0◦+9.2◦−3.7◦ from various
experimental and theoretical constraints [13].
12
1.7 Four-Momentum and Invariant Mass
In relativistic mechanics, some fundamental information about a particle is defined
by that particle’s four-momentum, pμ. Four-momentum is similar to non-relativistic
momentum, p, except that the energy of the particle is also one of the elements:
pμ = (E
c, px, py, pz), (1.18)
where E2 = (mc2)2 +(pc)2. The Minkowski inner product of pμ with itself has a very
simple result:
pμpμ =
(E
cpx py pz
)⎛⎜⎜⎜⎜⎜⎜⎜⎝
1 0 0 0
0 −1 0 0
0 0 −1 0
0 0 0 −1
⎞⎟⎟⎟⎟⎟⎟⎟⎠
⎛⎜⎜⎜⎜⎜⎜⎜⎝
Ec
px
py
pz
⎞⎟⎟⎟⎟⎟⎟⎟⎠
=E2
c2− p2 = m2c2. (1.19)
Thus, if we can know the four-momentum of a particle, we will know the particle’s
mass. Since momentum is conserved, we can use these principles to find the masses
of particles which we cannot observe. For example, consider the decay φ → K+K−.
Since momentum is conserved, pμφ = pμ
K+ + pμK− . We may not be able to see the
φ with a detector because it decays too quickly, but suppose we have found two
oppositely charged kaons and we are curious about their origin. We can determine
(pK+ + pK−)μ(pK+ + pK−)μ = m2K+K−c2 and define mK+K− as the invariant mass
of the two kaon system. If these kaons truly came from a φ, we must find that
mK+K− = mφ, since pφμpφμ ≡ m2
φc2.
13
1.8 Resonances and Interference
The decay D0 → K+K−π0 may occur directly, with one particle (the D0) decaying
into three particles without any intermediate states. However, the D0 may also decay
into two particles, followed by a subsequent two-body decay of one of those particles.
The three possibilities are: D0 → K+R−1 , R−
1 → K−π0; D0 → K−R+2 , R+
2 → K+π0;
and D0 → π0R03, R0
3 → K+K−. Ri represents a short-lived intermediate particle
which is called a resonance.
Quantum mechanical interference occurs when a physical process has multiple in-
distinguishable paths to completion. If there could be two or more different processes
for D0 → K+K−π0, and we can not distinguish which process participated in the
decay, we expect to observe interference between the processes.
1.9 The Dalitz Plot Formalism
The D0, K+, K−, and π0 mesons are all spin zero particles. In studying the decay
D0 → K+K−π0, it is important to consider how many degrees of freedom are required
to completely describe the kinematics. In the spin zero D-meson rest frame, there are
three decay particles each with four degrees of freedom (their four-momenta) for a
total of 12 unknowns. There is one constraint for each daughter particle mass (known
by experiment) and four additional constraints from the conservation of momentum
and energy in the decay. Finally, the three degrees of freedom describing the spatial
orientation of the decay are irrelevant since we have a spin zero D-meson. Thus, only
two independent degrees of freedom remain from the original twelve.
Since four-momentum is conserved during our decay, we know that:
pμD0 = pμ
K+ + pμK− + pμ
π0 . (1.20)
14
The Minkowski inner products of each side of Equation 1.20 with themselves should
also be equal. Such an operation, along with some rearranging, results in an equation
in terms of four constants and three invariant masses:
m2D0 + m2
K+ + m2K− + m2
π0 = m2K−π0 + m2
K+π0 + m2K+K− , (1.21)
where we define the invariant mass mab for particles a and b as:
mab =
√(pa + pb)μ(pa + pb)μ
c2. (1.22)
It is possible to form three possible two-particle invariant mass-squared variables
in a three-body decay. In our case, these are the mass-squared variables m2K+π0 ,
m2K−π0 , and m2
K+K− shown in Equation 1.21. These three invariant-mass squared
variables can be measured for each D0 → K+K−π0 candidate. Since the left side of
Equation 1.21 is a constant, only two of the invariant mass-squared variables on the
right side of the equation are independent. Hence all unique information about an
event is contained in any one of the three possible pairs of these variables.
A scatter plot of two of the mass-squared variables for a series of events is referred
to as a Dalitz plot [14] and can give us insight into the internal structure of the decay.
For example, suppose D0 → K+R−1 and then R−
1 → K−π0. Then we would see events
along the m2K−π0 axis of the Dalitz plot with the mass of the R−
1 resonance squared,
m2R−
1
. Meanwhile, the profile of the band along the orthogonal variable tells us about
its spin (this will be discussed further in Chapter 2, Section 2.3). A Dalitz plot
contains a wealth of information about the substructure of three-body decays (as well
as being a beautiful example of easily observable quantum-mechanical interference).
When several bands are present, their relative size and their interference in regions
of overlap tell us about the relative complex amplitudes of the resonances. Thus,
15
looking at the Dalitz plot gives us an idea of what resonances the decay may have
gone through and how they interfere.
The kinematically-allowed range of invariant mass-squared values define all of the
possible states (or “phase space”) of the system. Choosing a pair of invariant mass-
squared variables (rather than invariant mass, for example) as our degrees of freedom
is advantageous, as the decay fraction for any point in the phase space defined by
the two chosen invariant mass-squared variables is proportional to the square of the
matrix element [15], “M ” ≡ 〈K+K−π0|M|D0〉, describing the decay amplitude:
dΓ =|M|2
256π3m3D
dm2abdm2
bc, (1.23)
where a, b, and c could be any cyclic permutation of {K+, K−, and π0}. Note that in
the absence of intermediate resonances (which means a featureless matrix element),
scatter plots of any two invariant mass-squared variables will yield uniform distribu-
tions. It is also very advantageous that the relativistic invariance of the invariant
masses means that the Lorentz frame in which they are evaluated is irrelevant.
16
2 Model for Three-body Decays
The general amplitude for a D0 decay to a particle c and a resonance r, where r then
decays to particles a and b, and a, b, and c are pseudo-scalars (spin 0), is [15]:
AD0→(ab)rc =∑
λ
〈ab|rλ〉 Tr(mab) 〈crλ|D0〉
= Zl BD0
L BrL Tr(mab), (2.1)
where the sum is over the helicity states λ of r, L is the orbital angular momentum
between r and c, l is the orbital angular momentum between a and b (the spin of r),
Zl describes the angular distribution of the final state particles, BD0
L and BrL are the
barrier factors for the production of rc and of ab, and Tr is the dynamical function
describing the resonance r. There are multiple choices for the modeling of Zl, BL, and
Tr, which vary by experiment and the required level of detail necessary to describe a
particular three-body decay. Usually the resonances are modeled with a Breit-Wigner
formalism as will be discussed in this chapter. However, some recent analyses have
been done using a K-matrix formalism [16] with the P -vector approximation [17]. The
K-matrix formalism and reasons for considering it are discussed briefly in Appendix
A.
2.1 Breit-Wigner Formalism
The amplitude for a D0 decay to a particle c and a resonance r, where r then decays to
particles a and b, may be derived from the Breit-Wigner propagator and the Feynman
17
D0
c a
b
ab resonanceF
DF
rε
λ ελ
ελ
ε
λ
∗
λ
Σ
Figure 2.1: D0 → abc through an ab resonance. The initial and final states both haveno angular momentum, so the intermediate spin states are summed over.
rules, and is defined as [14]:
AD0→(ab)rc = (pD0 − pc)μ(Breit-Wigner propagator)μν(pa − pb)ν
= (pd − pc)μ
∑λ
εμ∗λ εν
λ
m2r − m2
ab − imrΓ(mab)(pa − pb)ν . (2.2)
This decay is represented in Figure 2.1. The amplitude in Equation 2.2 only
applies to point particles. In order to account for the finite size of our particles, we
add decay form factors at each vertex, FD and Fr:
AD0→(ab)rc = FD(mab)(pd − pc)μ
∑λ
εμ∗λ εν
λ
m2r − m2
ab − imrΓ(mab)(pa − pb)νFr(mab). (2.3)
These form factors are unknown, but in practice they are set to the Blatt-Weisskopf
penetration factors [18] as shown in Table 2.1, where rD0 or rr is the “radial parame-
ter” of the decaying meson (D0 or r, respectively). We choose to have rD0 = 5 hcGeV
for
the D0 and rr = 1.5 hcGeV
for the intermediate resonance. These values are consistent
with previous Dalitz plot analyses [4, 14, 19], and small changes in rD0 and rr were
shown in those analyses to have a very small effect on the fit results.
18
Spin FP (mab)
0 1
1
√1 + zP (mr)1 + zP (mab)
2
√9 + 3zP (mr) + z2
P (mr)9 + 3zP (mab) + z2
P (mab)
Table 2.1: Normalized Blatt-Weisskopf Barrier Factors (FP (mab) = 1 when mab = mr)for the particle P decay vertex. zP is defined as follows: zP (mab) = r2
P p2P (mab).
Γ(mab) in the denominator of Equation 2.3 is the mass-dependent width [20] and
depends on the spin of the resonance:
Γ(J)(mab) = Γrmr
mab
(pr(mab)
pr(mr)
)2J+1
F (J)r
2. (2.4)
FD, Fr, and Γ(mab) are dependent on the breakup momentum of the related
particle D or r at invariant mass mab. It helps to recall the general expression for the
breakup momentum, pP , which is the magnitude of the momentum of either daughter
(masses m1 and m2) from the two-body decay of a particle P (mass m12) in the P
rest frame:
pP =
√(m2
12 − (m1 + m2)2)(m212 − (m1 − m2)2)
4m212
. (2.5)
We can compare Equations 2.1 and 2.3 to see how each of the parts of Equation
19
2.1 is modeled:
Zl = (pd − pc)μ
∑λ
εμ∗λ εν
λ(pa − pb)ν ,
BD0
L = F(J)D (mab),
BrL = F (J)
r (mab),
and Tr(mab) =1
m2r − m2
ab − imrΓ(mab). (2.6)
2.2 Decay Amplitudes for a Particular Spin
Equation 2.3 may be evaluated for resonances having different spins. For a scalar
(spin 0) resonance the result is:
Ascalar =1
m2r − m2
ab − imrΓ(0)(m2ab)
. (2.7)
For a vector (spin 1) resonance, the spin sum in the numerator of Equation 2.3 is
evaluated to [21]:
−gμν +pμ
abpνab
mab
. (2.8)
With this information, we can evaluate Equation 2.3 for a vector resonance as:
Avector =F
(1)D (m2
ab) · F (1)r (m2
ab)
m2r − m2
ab − imrΓ(1)(m2ab)
·[m2
ca − m2bc +
(m2D − m2
c)(m2b − m2
a)
m2ab
]. (2.9)
The spin 2 resonance amplitude is also used in our analysis. A partial derivation
20
is in reference [14]. The result of this derivation is the following amplitude:
Atensor =F
(2)D (m2
ab) · F (2)r (m2
ab)
m2r − m2
ab − imrΓ(2)(m2
ab)·⎡
⎢⎣ −13(m2
ab − 2m2D − 2m2
c +(m2
D−m2c)2
m2ab
)(m2ab − 2m2
a − 2m2b +
(m2a−m2
b)2
m2ab
)
+(m2bc − m2
ca +(m2
D−m2c)(m2
a−m2b)
m2ab
)2
⎤⎥⎦ . (2.10)
Higher spin intermediate states were not necessary for our analysis. The non-
resonant amplitude is simply a constant (it doesn’t change at different points on the
Dalitz plot).
In our analysis, the parametrization of resonances which decay r → bc or r →ca (in m2
bc and m2ca invariant mass squared variables, respectively) are obtained
from Equations 2.7, 2.9, and 2.10 by simple cyclic permutations of the {a, b, c} =
{K+, K−, π0} particles and {m2ab,m
2bc,m
2ca} = {m2
K+K− ,m2K−π0 ,m2
π0K+} variables.
This is an important specification to make. Note that swapping a and b in Equa-
tion 2.9 will result in the amplitude being multiplied by −1 (which is equivalent to
a phase shift of 180◦). Thus, we resolve any ambiguities by specifying that we use
cyclic permutations when defining amplitudes. However, the order of the particles
that we choose to cyclically permute is also important. For example, if we choose
permutations of {K−, K+, π0} instead of {K+, K−, π0} in our definitions of the vec-
tor resonances, then each vector resonance will have a 180◦ phase shift. This would
affect the interference between vector resonances and those contributions which do
not have a phase shift based on particle order (such as the non-resonant contribution).
When comparing Dalitz plot fits from different experiments, it is crucial to account
for differences in phase conventions.
21
D0
K-
J = 1J = 0 J = 0
θπ
0
K+
φ
Figure 2.2: Since the D0 is a spinless particle, a vector (spin 1) decay particle suchas the φ must be in a |1, 0〉 angular momentum state in the D0 rest frame (when wequantize along the π0 direction). The daughters of the φ therefore decay preferentiallyparallel to its momentum (the amplitude has a Y 1
0 ∝ cos(θ) dependence in the D0
rest frame).
2.3 Amplitudes on the Dalitz Plot
Although not obvious from looking at Equation 2.9, this amplitude is proportional
to cos(θ) in the D0 rest frame, where θ is the angle between the momentum vector
of the resonance and the momentum vector of one of the daughters as shown in
Figure 2.2. However, this should not be surprising because the D0 has no spin and
the resonance has spin 1, therefore the angular dependence in the decay amplitude
should be represented by the Y 01 spherical harmonic, which is proportional to cos(θ).
This causes an interesting feature in the Dalitz plot variables orthogonal to the one
in which the resonance decays. Because of the cos(θ) dependence of the amplitude,
we will find less events near the center of the resonant band in the orthogonal Dalitz
plot variable and more events near the edges, as seen in Figure 2.3. Similarly, we
expect scalar (spin 0) resonances to have no preferred angular direction (Y 00 ), and
tensor (spin 2) resonances to have an angular dependence in the decay amplitude
represented by Y 02 ∝ 3 cos2(θ) − 1.
22
0.5
1.5
1.0
1.0
0.5
1.5
mK -
π 02 (GeV 2/c 4) m
K+ π
02 (G
eV2 /c
4 )
θ = 180°
θ = 90°
θ = 0°
mK*
2
Figure 2.3: |A|2 for the vector K−π0 resonance K∗−. When the π0 comes out oppositeto the K∗− direction, it is almost at rest and corresponds to θ = 0◦. The K− is almostat rest when θ = 180◦. When the K− and π0 come out perpendicular to the K∗−
(θ = 90◦), they have equal momentum. We can see the preference for a decay tooccur parallel to the momentum of the K∗− as a cos2(θ) dependence on the Dalitzplot (the square of the cos(θ) amplitude dependence mentioned in Figure 2.2).
23
3 CLEO, CESR, and LEPP
We obtain the data for our analysis using the Cornell Electron Storage Ring (CESR)
and CLEO detector at the Laboratory for Elementary-Particle Physics (LEPP).
LEPP is located on the Cornell University campus in Ithaca, NY. There is one inter-
action region at the south end of CESR which houses the CLEO detector. The CLEO
detector first took data in 1979, and has been upgraded on numerous occasions. For
our analyses, we have used data collected by the CLEO III [22–24] and CLEO-c [25]
versions of the detector.
3.1 LEPP
The main mission of LEPP is to study nature’s fundamental particles and the laws
that govern them. LEPP is also responsible for the development of technology behind
accelerators which are a major tool in learning about particle physics. The CLEO
experiment and CESR are integral to fulfilling LEPP’s charge. In addition to its
role in particle physics, LEPP is also responsible for providing x-rays to scientists via
the Cornell High Energy Synchrotron Source (CHESS), which harnesses synchrotron
radiation from the particles which are stored in CESR. This intense radiation, in the
form of x-rays, is used by surface physicists, medical biologists, and others to study
the microscopic structure of many materials.
24
e-
Transfer Linee+
Transfer Line
LINAC
Converter
CLEO
Synchrotron
CESR
Gun
e+
e-
Figure 3.1: A diagram of CESR.
3.2 CESR
CESR (Figure 3.1) is a symmetric e+e− collider capable of running at center-of-mass
energies between approximately 3 and 11 GeV. While “CESR” can be taken to refer
specifically to the storage ring, it can also refer to the entire apparatus used to create
and accelerate positrons and electrons. The entire apparatus consists mainly of a
linear accelerator (LINAC), synchrotron, and the storage ring. The storage ring
stores both electrons and positrons and has a circumference of 768 m. It is located
underneath the Robison Alumni Fields, the Robert J. Kane Sports Complex, the
Friedman Wrestling Center, and the Kite Hill parking lot.
Electrons are produced from a heated filament in an electron gun, and the elec-
trons are collected in a “prebuncher” which compresses electrons into packets. The
electron packets are accelerated in the LINAC using varying electric fields generated
by radio frequency (RF) cavities. The electrons have an energy of approximately 300
MeV at the end of the 30 m LINAC. About half-way (15 m) down the LINAC, the
electron beam may collide with a movable tungsten target. Collisions with the tung-
25
sten target create showers of many particles which include positrons. The positrons
are separated from the other particles (mostly electrons, x-ray photons, and protons)
and accelerated in the remainder of the LINAC to an energy of about 150 MeV.
Bunches of electrons and positrons from the LINAC are injected in opposite di-
rections into the synchrotron. The synchrotron is a few meters smaller in radius than
the storage ring and is located in the same tunnel. In the synchrotron, the particles
are accelerated by four 3-meter long linear accelerators, and contained by a series
of dipole bending magnets. The magnetic fields are increased as the energy of the
particles is increased in order to continuously contain the particles. After reaching
their final energy, the particles are transferred to the storage ring.
In the storage ring, electrons and positrons are guided by dipole bending magnets
and are focused by a series of quadrupole and sextupole magnets. The beams lose
energy by synchrotron radiation which occurs as charged particles move in a curved
path, so superconducting RF cavities are used to maintain the beam energy, thus
keeping the particles in their orbits. The electrons and positrons are stored in the
same vacuum beam pipe, which means that care must be taken to avoid unwanted col-
lisions. Electrostatic separators are used to bend the paths of electrons and positrons
into trajectories which only intersect in the interaction region. This configuration is
known as a “pretzel”-shaped orbit, although in reality the deviations of the electrons
and positrons are minimal. For low-energy running (center-of-mass energies near 4
GeV), “wiggler” magnets induce synchrotron radiation, but this has a minimal effect
on the beam trajectory.
The electrons and positrons are not continually placed into the storage ring, but
are rather located in “bunches”, which are grouped into “trains”. CESR can be
configured to store up to 9 trains with a maximum of 5 bunches each. The bunches
in the trains are separated by 14 ns, and the trains themselves are separated by 284 ns.
When an electron and a positron collide, they either annihilate, producing a virtual
26
Solenoid Coil BarrelCalorimeter
Drift Chamber
Silicon Vertex Detector / Beampipe
EndcapCalorimeter
IronPolepiece
Barrel MuonChambers
MagnetIron
Rare EarthQuadrupole
SCQuadrupoles
SC QuadrupolePylon
Ring Imaging Cherenkov Detector
Figure 3.2: The CLEO III Detector.
photon which decays into a pair of fermions, or scatter. The possible decay fermions
depend on the beam energy, but are always produced in a matter-antimatter pair.
It is also possible that the initial electron and positron radiate two photons, which
subsequently collide. For our analyses, we are mainly interested in D0’s and D0’s
produced from cc pairs generated in continuum at or near the Υ(4S) center-of-mass
energy (CLEO III) or generated at the ψ(3770) center-of-mass energy (CLEO-c).
3.3 The CLEO III Detector
The CLEO III detector shown in Figure 3.2 consists of (from the beam line outwards)
a four-layer silicon strip vertex detector, a 47-layer wire drift chamber using a helium-
propane gas mixture (60:40), a ring imaging Cherenkov (RICH) detector [26], an
electromagnetic crystal calorimeter (made of cesium iodide doped with thallium), a
superconducting solenoid maintaining a 1.5 T magnetic field parallel to the beam line,
27
Figure 3.3: End view and quarter-section view of the CLEO III Silicon Detector.
and a muon chamber system [22, 25]. For the particles in which we are interested,
charged mesons and π0 mesons, the most interesting parts of the detector are the
tracking system, the RICH, and the calorimeter.
3.3.1 Silicon Vertex Detector
The tracking system, consisting of the silicon vertex detector and the drift chamber,
covers 93% of the 4π solid angle around the interaction region. The silicon vertex
detector (Figure 3.3) provides accurate track position measurements close to the
interaction point in perpendicular distance from beam line (r), azimuthal angle (φ),
and parallel distance along beam line (z) coordinates. It consists of four 300 μm thick
detection layers which circle around the beam line at 2.5, 3.76, 7.0, and 10.1 cm. Each
detection layer is built of 5.0 cm × 2.5 cm silicon detectors. Each individual detector
has silicon strips every 50 μm along r − φ (on the obverse side) and 100 μm along
z (on the reverse side). Each layer consists of a proportionally increasing number
28
circle of closestapproach
14 mm
electronsdrift and multiply
sense wirefield wire
char
ged
part
icle
traj
ecto
ry
electron
Figure 3.4: Charge collection and multiplication in the drift chamber.
of detectors along φ and z (starting with 7 along φ and 3 along z in the 2.5 cm
layer); a total of 447 silicon detectors are used to make the four layers. Precision
information about the direction of the pion in D∗ → Dπ decays allows for a D∗ − D
mass difference resolution as good as 0.19 MeV/c2 [25,27].
3.3.2 Drift Chamber
The drift chamber covers a region in r of 13.2 cm to 82.0 cm from the beamline. The
inner wall of the drift chamber is 2 mm thick expanded acrylic with 20 μm aluminum
skins. The outer radial wall is made up of two layers of 0.8 mm thick aluminum
cylindrical shells. The drift chamber is made up of 9796 drift cells and is filled with a
60% helium, 40% propane gas mixture. At the center of each cell is a 20 μm diameter
gold-plated tungsten sense wire, and the sense wire is surrounded by eight 110 μm
diameter gold-plated aluminum field wires. The cell forms a nearly square shape 14
mm across. A 2100 V potential difference is applied between the sense wire and the
field wires, creating an electric field.
29
back FRONT
a track pointing toward the FRONT of the detectorwire projection
z
stereo anglesexaggerated
stereo wire
s tilted counter−clockwise
stereo wires tilte
d clockwise
axial wires (untilted)
FRONTview
Figure 3.5: Using tilted stereo wires to obtain z information in the drift chamber.Tilted wires are projected as arrows coming out of the page in the r-φ projection;position along a tilted wire indicates the z of the track helix near that wire. Theclosest wires to the track (in three dimensions) are highlighted (in green).
A charged particle will ionize the gas mixture when it passes through a cell (see
Figure 3.4), and the ionized electrons are then attracted to the sense wire. The
electric field near the sense wire is very strong and causes the ionized electrons to
ionize more atoms. This creates a avalanche of electrons on the sense wire. A distance
of closest approach to the sense wire can be calculated based on the transit time of
the electron pulse from the wire, and the timing of collisions in the detector. The
wire which carried the pulse is noted, and is used in concert with information from
other wires to help determine the trajectory of the charged particle.
The drift chamber consists of 16 layers of axial sense wires (parallel to the beam-
line) and 31 sense wire layers which alternate in small stereo angles to provide some
information about the position in z of the tracks. The wires in a stereo layer are
skewed in the φ direction so that the φ coordinate at one end of the wire is offset by
a small angle from the other end (see Figure 3.5). The stereo angle varies from 21 to
30
28 mrad with respect to the beam axis and alternate in each subsequent layer.
The outer radial wall is lined with 1 cm wide cathode pads. The cathode pads
provide a longitudinal position measurement at the outer radius. The cathodes, when
used with the stereo layers, improve the spatial measurement in the z direction. The
combined z resolution is on the order of a couple millimeters.
Overall, the drift chamber has an average position resolution of 100 μm and
momentum resolution of 40 MeV/c at p = 5.3 GeV/c. [25, 27].
3.3.3 dE/dx Particle Identification
The energy loss per unit length, dE/dx, and momentum of a particle are correlated
for different charged particles which may leave a track in the detector. The nature
of this energy loss is the ionization or atomic excitation of moderately relativistic
particles traveling through matter. At the energies of the CLEO experiment, dE/dx
is a function of only the particle’s velocity, as determined by Bethe and Bloch [1].
Together with a measurement of the particle’s momentum, the mass of the particle
can be determined from the velocity and momentum as follows:
m = p
√1
v2− 1
c2. (3.1)
Figure 3.6 shows how different charged hadrons have different correlations between
their momentum and energy loss. Notice that in this energy range dE/dx ∝ 1v2 , and
thus the curve has a relationship to the inverse square of the momentum as well. We
can use our knowledge of these correlations to perform particle identification with the
dE/dx information.
If one had a particular track’s momentum and the dE/dx information it is possible
to determine how many standard deviations away from a particular particle type the
track’s dE/dx information lies. It is clear from Figure 3.6 that the effectiveness of
31
p (GeV/c)
dE/
dx
0
1
2
3
4
5
0 0.25 0.5 0.75 1 1.25 1.5
π
K
p
Figure 3.6: CLEO III dE/dx vs. momentum showing the π, K and p bands. Leptonsare suppressed where possible. The dE/dx scale is chosen to be 1 for electrons andpositrons from e+e− → e+e− events.
32
using dE/dx information to determine type of particle is dependent on momentum.
Areas of overlap in Figure 3.6 are regions where the ability to distinguish particle
type is lost. Figure 3.6 shows that distinguishing kaons and pions with dE/dx starts
to become effective at momenta below approximately 700 MeV/c. If the number of
standard deviations, σ, for a given particle type is 3 or less (in the lower momentum
region where we can distinguish kaons and pions), a positive identification may be
made.
Typically dE/dx information can be used to distinguish two types of particles
(the π and the K for example) by looking at the χ2 difference Δχ2dE/dx:
Δχ2dE/dx = σ2
K − σ2π, (3.2)
where particles with Δχ2dE/dx < 0 are more likely to be kaons than pions.
3.3.4 RICH Detector
The RICH detector (Figure 3.7) is located between the drift chamber and the calorime-
ter and covers 83% of 4π. Charged particles passing through 10 mm thick lithium
fluoride crystals create a ring of Cherenkov photons that expands by traveling through
a nitrogen expansion gap which is 16 cm in length. For the region of the RICH de-
tector which surrounds the interaction region, the interface of the radiating lithium
fluoride crystals and the nitrogen is shaped as a sawtooth rather than flat in order
to prevent total internal reflection of Cherenkov light from incident tracks normal
to the boundary. Further from the interaction area, total internal reflection is not a
problem, and flat crystals are used. After traveling through the expansion gap, the
Cherenkov photons are detected via conversion into photo-electrons by interaction
with a methane-triethylamine mixture. These photo-electrons are multiplied by a
multi-wire proportional chamber mechanism. The position of the photon conversion
33
Methane - TEA
CaF2
Windows
N2 Expansion
Gap
K / π
LiF Radiator
Drif
t Cha
mb
er
θc
Figure 3.7: A section of the RICH detector.
34
is determined by induced charge on cathode pads at the end of the wire chamber [25].
The RICH is an important tool in distinguishing charged kaons (K±) from charged
pions (π±), because the mass of a charged K± or π± can be inferred from the opening
angle of the ring of photons and the particle’s momentum. Cherenkov radiation is
produced by the charged particles in the lithium fluoride crystal at an angle θc to the
trajectory of the track. The angle depends on the velocity of the particle, v, and the
index of refraction, nLiF = 1.5, of the lithium fluoride as follows:
cos θc =c
vnLiF
. (3.3)
Of course, it follows that the minimum velocity that a particle must have in order
to radiate in the detector is approximately 23c. The (rest) mass m of the particle can
then be determined from θc and the measured magnitude of momentum p, since the
relativistic momentum of the particle is given by:
p =mv√1 − v2
c2
. (3.4)
Information from the RICH is used to determine a likelihood for a particular
particle hypothesis. A likelihood L for each particle type is calculated from the
number of photons which are within 5 standard deviations of the expected ring size
for that particle type. An effective χ2 for that particular particle type is -2 ln(L).
Thus, we can separate particles in a familiar way by using:
Δχ2RICH = −2 ln(LK) − (−2 ln(Lπ)), (3.5)
where particles with Δχ2RICH < 0 are more likely to be kaons than pions.
It is important to note that tracks with a higher number of associated RICH
photons will have a higher quality of likelihood determination. This is because the
35
0 1 2 3 4Momentum (GeV)
-400
-200
0
Δχ
2 dE/
dx
0 1 2 3 4Momentum (GeV)
-400
-200
0
(b)
(a)
Δχ
2 RIC
H
Figure 3.8: How true kaons (simulated) would look at various momenta (binned) interms of the range of possible values for Δχ2
dE/dx (Equation 3.2) for dE/dx information
and the range of possible values for Δχ2RICH (Equation 3.5) for RICH information,
as defined in the text. Values of less then zero indicate that the kaon hypothesis ismore likely.
number of photons, Nγ, is determined within a 3 standard deviation band about
the expected radius, even though the likelihood is determined within a 5 standard
deviation band about the expected radius. So it is possible to have likelihoods which
are non-zero, even when Nγ is zero! Typically, choosing tracks with Nγ ≥ 3 is
necessary for reliable particle identification. Typical selection requirements for the
RICH are able to identify 92% of kaons with a pion fake rate of 8%1 [25].
The particle ID capabilities of dE/dx and the RICH fortunately compliment each
other. In the case of K/π separation, RICH information is not good below a mo-
mentum of about 700 MeV/c because the kaons do not radiate in the RICH at lower
1This is for a sample of tracks from D0 → K−π+ decays (produced via D∗+ → D0π+ decays)with momenta in between 0.7 and 2.7 GeV/c.
36
momenta. The dE/dx separation is good below about 700 MeV/c, but is not very
useful above 700 MeV/c. Figure 3.8 shows the ability of each particle identification
system to identify kaons at various momenta.
RICH and dE/dx information may be combined into a single overall chi-squared
difference:
Δχ2 = −2 ln(LK) − (−2 ln(Lπ)) + σ2K − σ2
π, (3.6)
where particles with Δχ2 < 0 are more likely to be kaons.
3.3.5 Crystal Calorimeter
The 7800-crystal thallium-doped cesium iodide electromagnetic calorimeter is used
mainly to help find photons and identify electrons, and covers 93% of the 4π solid
angle. A π0 decays to two photons about 98.8% of the time [1], so when looking for
a π0 we aim to reconstruct it from these two photons. These photons interact with
the calorimeter creating showers of charged particles and additional photons. The
scintillation light from these showers is detected by silicon photodiodes located on the
back of each crystal, and the signal is used to determine the energy of the photon.
The face of each crystal is a 5 by 5 cm square, and the length of each crystal is 30 cm.
The crystals are oriented to point approximately towards the interaction region. The
calorimeter has a mass resolution for π0 → γγ of approximately 6 MeV/c2 depending
on photon energies and locations [25]. The calorimeter is optimized to capture all
of the energy of electrons and photons which interact with it, as the length of each
crystal is approximately 16 radiation lengths.
37
3.3.6 Solenoid Coil
The solenoid coil is a large superconducting magnet that is cooled by liquid helium in
order to produce a strong, uniform magnetic field of 1.5 T. This field causes charged
particles to make curved paths in the detector. The degree and the direction of a
particle’s curvature can be used to help identify the particle. The coil encompasses
all of the detector elements except for the muon detectors.
3.3.7 Muon Detectors
The muon detector system consists of three “superlayers” which contain many plastic
tubes surrounding anode wires. When a charged particle passes through the muon
detector, an electrical signal is generated in a similar manner to those signals gener-
ated in the drift chamber. There are layers of iron which serve to stop most particles
which would otherwise escape the detector. Muons are able to penetrate the iron,
and the depth to which the muon travels helps to identify it. About 85% of the
solid angle is covered by the muon detectors. This analysis does not use any of the
information provided by the muon detector.
3.3.8 Trigger and Data Acquisition Systems
Events are recorded by the Data Acquisition System (DAQ). All events cannot be
recorded because they happen at a rate which is much too fast to record. Also, many
events are relatively uninteresting, and thus not worth recording. A relatively large
amount of time is required to reconstruct the event and write the event information
from the detector to disk, so only events that contain interesting physics are recorded.
The triggering system designed and built by the University of Illinois performs
pattern recognition algorithms which determine which events are good candidates for
recording. After information from the detector becomes available, the trigger chooses
38
1 crystal
Figure 3.9: How the trigger tiles read shower information from the calorimeter.
to store an event based on multiple selection criteria. The trigger uses information
from tracks in the drift chamber and showers in the calorimeter to make its decisions
without fully reconstructing the event. The tracking trigger is further broken down
into axial and stereo components.
The axial layers have 1696 wires and information from each wire is read out
individually. There are 8100 wires in the stereo layers, so in order to get information
more quickly, the wires are read out in groups of 16 wires in 4× 4 wire groups called
“blocks”. Each block is read out to the trigger as a single piece of information, which
represents a wire hit or lack thereof. The information from both the stereo and axial
trigger is correlated into a proto-track, which is fed into the Level 1 (global) trigger.
Also fed into the trigger is information from the calorimeter. The trigger basically
looks for how much energy is deposited in the calorimeter, and where. However, a
low energy shower that is split across multiple crystals may not have enough energy
39
to surpass the threshold for what would constitute a hit in the calorimeter. In order
to overcome this difficulty, the trigger reads out the energy from the calorimeter in
groups of 64 crystals called “tiles”. There is overlap in the tiles, such that a signal
in one crystal will appear in four different tiles, as seen in Figure 3.9. At least one
tile will contain the majority of a shower. A signal in a single crystal will show up in
four tiles, and this is accounted for by the calorimeter trigger.
The Level 1 trigger looks at these tracks and showers and determine if the event is
interesting. The Level 1 Trigger looks for sets of tracks and showers which correspond
to predefined categories. These categories, or “trigger lines”, each look for basic event
properties to determine if the event should be recorded.
If a particular trigger line is satisfied, the name of this line is recorded so that
events can later be sorted into groups, making it easy to access general event categories
of interest for analysis.
Should the event pass the trigger, all of the event information from the detector
is recorded to the DAQ. The DAQ takes about 30 μs to record information about an
event, and can do so at a rate of 500 Hz. A typical data size for a single event is
25 kB and the data transfer bandwidth is 6 MB/s. The average detector deadtime
(time when the detector cannot collect any information) is 3%.
3.4 The CLEO-c Detector
The CLEO-c Detector (Figure 3.10) is basically identical to the CLEO-III detector,
except that the magnetic field is lowered from 1.5T to 1.0T and the silicon vertex
detector is replaced with a small inner wire chamber called the ZD (Figure 3.11).
The purpose of these changes is to help in the detection of particles with significantly
lower momentum, due to a decrease in running energy for the experiment from the
(CLEO-III) energy range of the Υ resonance family to the (CLEO-c) energy range of
40
CsIElectromagnetic
Calorimeter
(End Cap)
CsIElectromagnetic
Calorimeter
Drift ChamberOuter Endplate
Drift Chamber EndplateSmall Radius Section
Drift ChamberInner Skin
cos θ = 0.93
Interaction Point
CESR
ZD Inner DriftChamber
Ring Imaging CherenkovDetector
Figure 3.10: Cross section of the CLEO-c detector in the r-z plane.
41
Figure 3.11: An isometric view of the CLEO-c inner wire chamber.
the J/ψ resonance family.
3.4.1 The ZD
The ZD is a six-layer inner drift chamber which is located between the beam pipe
and the main (CLEO-III) drift chamber. It consists of 300 drift cells and is filled
with a 60% helium, 40% propane gas mixture. At the center of each cell is a sense
wire, and the sense wire is surrounded by eight field wires. The wires are made of the
same materials as those in the main drift chamber. The cell forms a nearly square
shape 10 mm across. A 1900 V potential difference is applied between the sense wire
and the field wires. All of the layers in the ZD are arranged at stereo angles, with
the innermost layer being at a stereo angle of 4.4◦ with respect to the beam axis and
the outermost layer being at a stereo angle of 5.8◦ with respect to the beam axis.
The process of charged particle detection is identical to the process of detection in
the main drift chamber.
The ZD covers the region which is between 4.1 and 11.7 cm from the beamline.
42
The inner wall is made of aluminum which is 1 mm thick, and the outer radial wall
is made of mylar which is 127 μm thick. The resolution in the position of charged
particles along the z direction is 680 μm. The ZD has a momentum resolution, σp
p, of
about 0.4% for charged particles at normal incidence. The ZD is the only source of
z information for a charged particle with transverse momentum less than 67 MeV/c.
3.5 Reconstructing Events
After an event is read out, the next step is to process the data into a form useful
for analysis. This analysis process is performed by an off-line computer code called
“Pass2”, which performs the full reconstruction and fitting of the tracks, and the clus-
tering of calorimeter showers (there is also a code called “Pass1” which is performed
on the data as it is taken, but this is mostly to ensure the quality of the data).
The first step is to determine the set of calibration constants for the events, which
help convert the raw information from the detector into meaningful physical qualities,
and which can also help remove noise from the detector. After this step, higher level
reconstruction occurs. This includes building the tracks and showers, and matching
them. Quantities which are used for identifying tracks, such as dE/dx and RICH
information, are also calculated.
Pass2 also reconstructs short lived particles such as the π0 which decay into two
photons that are only detected in the calorimeter. Showers from the calorimeter are
combined to determine if the showers had the right energy to be a π0. A list of these
particles is created for data analysis.
43
3.6 Monte Carlo Simulation
In order to prevent bias due to selection criteria, examine sources of background, and
determine detector efficiency, particle experimentalists need the ability to create a
simulated data sample, which we call “Monte Carlo” (MC).
Creation of Monte Carlo is done in multiple stages. The first stage, using the
simulation program QQ [28], simulates the e+ - e− collision and primary decays of
the particles at the particular beam energy and initial conditions of the beam. The
Monte Carlo uses a decay file that gives the probability of each possible decay of a
particle, and one of the decay chains is randomly selected based on these probabilities.
The output of QQ is the four-momentum of each daughter particle. The next
simulation program, GEANT [29], takes the QQ particles and runs them through
a simulation of the detector. All of the interactions of the particles in the detector
are simulated, including bremsstrahlung radiation and interactions with material.
GEANT also accounts for resolution effects, detector efficiencies, and noise. GEANT
can be set to use the same calibration constants as any particular set of recorded
events, so that different detector settings may be used to generate MC samples in
order to represent all sets of events.
All of the simulated detector responses are put into a file which looks very much
like the one that is stored to the DAQ for real events, except that it also contains
information from QQ. This means that for simulated data, we can have the informa-
tion about which particles decayed and what they decayed into. With real data we
can only make hypotheses. The simulation can then be run through Pass2, just like
real events, and the MC can be analyzed in the same way as the data.
44
4 D0 → K+K−π0 in CLEO III
The main goal of the analysis which we describe here1 is to measure the strong phase
difference δD and relative amplitude rD between D0 → K∗−K+ and D0 → K∗+K−
decays, which is required for the extraction of γ via the method of Grossman, Ligeti,
and Soffer [10] as discussed in Section 1.6. We are further motivated by a paper by
Rosner and Suprun [30] that points out the sensitivity to δD using D0 → K+K−π0
produced in e+e− → ψ(3770) → D0D0
(CLEO-c), though the analysis presented here
relies on D0 mesons from D∗+ meson decays in e+e− continuum (non-BB) production
at√
s ≈ 10.58 GeV (CLEO-III).
4.1 Current Knowledge
This is the first analysis of the resonant substructures of D0 → K+K−π0 and their
interference. The only published measurement of the branching ratio (decay proba-
bility) was done by the CLEO collaboration in 1996 [31]. The current world average
for the branching ratio (BR) is BR(D0 → K+K−π0) = (0.13 ± 0.04%) [1]. Some
other published branching ratios of D0 decays which decay to our three-body mode
are shown in Table 4.1. For examples of possible Feynman diagrams for this decay,
see Figure 4.1.
1Some of the material in this chapter is our work previously published by the American PhysicalSociety (APS) in their journal, Physical Review D. The journal reference is: C. Cawlfield et al.(CLEO collaboration), Phys. Rev. D 74, 031108(R) (2006). Copyright to the article was transferredto APS. However, the transfer of copyright agreement states: “The author(s) shall have the followingrights:... The right, after publication by APS, to use all or part of the Article without revision ormodification, including the APS-formatted version, in personal compilations or other publicationsof the author’s own works...”
45
(a)
c
u_
s
u
u
u
u
s
_
_
_
D0
W+
K-
π0
K+
Nonresonant D0 K+ K-π0
(b)
c
u_
s
s
u
u_
u
s_
s_
D0
W+
K+
π0
φ
K-
D0 π0
K+ K-
(c)
c
u_
s
u_
u_
D0
W+
K-
K+
π0
u
s_ K*+
D0 K*+ K-
K+ π0
u
u
s_
(d)
c
u_
u_
ss
u
u
s
u_
u_
_
D0
W+
K-K*-
K+
π0
D0 K*- K +
K-π0
Color suppressed φ
g
gg
g
gg
g
g
u_
Figure 4.1: Shown are examples of (a) a non-resonant decay, (b) a neutral intermedi-ate state (φ), (c) a positively charged intermediate state (K∗+), and (d) a negativelycharged intermediate state (K∗−).
Decay Mode BR ReferenceD0 → K∗+K− (0.37 ± 0.08)% [1,32,33]D0 → K∗−K+ (0.20 ± 0.11)% [1,32]
D0 → φπ0 (0.076 ± 0.005)% [34]
Table 4.1: Branching ratios of intermediate D0 modes which may decay to the three-body mode K+K−π0.
46
Three-body decays of D mesons are expected to be dominated by resonant two-
body decays [35–39] and the well established Dalitz plot analysis technique [15] can
be used to explore their relative amplitudes and phases. The CLEO collaboration
has published Dalitz plot analyses for several three-body D0 decays over the past few
years [4,14,40–44] and the work described here closely follows the methods developed
in these previous analyses. The two most similar analyses in terms of technique are
CLEO analyses of D0 → K−π+π0 [14] and D0 → π−π+π0 [4, 40] decays.
4.2 Data Sample
This analysis uses an integrated luminosity2 of L = 9.0 fb−1 of e+e− collisions at
√s ≈ 10.58 GeV provided by the Cornell Electron Storage Ring (CESR). The data
were collected with the CLEO III detector described in Chapter 3.
To suppress backgrounds and to tag the flavor D0(D0), we require our D candidate
to come from a D∗+(cd) or D∗−(cd). The D0 mesons are reconstructed in the decay
sequence D∗+ → π+s D0, where the sign of the slow pion π+
s (π−s ) tags the flavor of the
D0(D0) at the time of its production. We call the charged pion ‘slow’ due to its small
expected momentum based on the relatively small mass difference between D∗+ and
D0.
The detected charged particle tracks must reconstruct to within 5 cm of the in-
teraction point along the beam pipe and within 5 mm perpendicular to the beam
pipe (the typical beam spot is 300 μm in the horizontal dimension, 100 μm in the
vertical dimension, and 10 mm in the longitudinal dimension). The cosine of the
angle between a track and the nominal beam axis must be between −0.9 and 0.9 in
2Instantaneous luminosity Linst. is a measure of the collision rate per unit area at the interactionpoint. The rate of a process with cross section σ is dN
dt = Linst.σ, where N is the number of timesthe process occurs. We may integrate the instantaneous luminosity over a period of time to get theintegrated luminosity L =
∫ Linst.dt. This way we may obtain N for a process by simply multiplyingthe integrated luminosity by the cross section for the process: N = Lσ.
47
order to assure that the particle is in the fiducial volume of the detector. The πs can-
didates are required to have momenta 150 ≤ pπs ≤ 500 MeV/c to remove particles
which are too slow to be properly detected and to remove pions which have too much
momentum to be good slow pion candidates. Kaon candidates are required to have
momenta 200 ≤ pK ≤ 5000 MeV/c.
Candidate kaon tracks that have momenta greater than or equal to 500 MeV/c
are selected based on information from the Ring Imaging Cherenkov (RICH) detector
if at least four photons associated with the track are detected. The pattern of the
Cherenkov photon hits in the RICH detector is fit to both a kaon and a pion hypoth-
esis, each with its own likelihood LK and Lπ. We require (−2 lnLK)− (−2 lnLπ) < 0
for a kaon candidate to be accepted. Candidate kaon tracks without RICH informa-
tion or with momentum below 500 MeV/c are required to have specific energy loss
in the drift chamber within 2.5 standard deviations of that expected for a true kaon
(|σK | < 2.5). Figure 4.2 shows plots of (−2 lnLK) − (−2 lnLπ), number of photons
nγ, and σK for the positive and negative kaons after all other selection criteria have
been applied.
The π0 candidates are reconstructed from all pairs of electromagnetic showers
that are not associated with charged tracks. To reduce the number of fake π0s from
random shower combinations, we require that each shower have an energy greater than
100 MeV and be in the barrel region of the detector. The two-photon invariant mass
is required to be within 2.5 standard deviations of the known π0 mass. To improve
the resolution on the π0 three-momentum, the γγ invariant mass is constrained to
the known π0 mass.
48
-250 -200 -150 -100 -50 0 500
100
200
300
400
500
[(- 2 ln LK) - (-2 ln L
π)]
K+
RICH
-250 -200 -150 -100 -50 0 500
100
200
300
400
[(- 2 ln LK) - (-2 ln L
π)]
K-
RICH
dE/dx
-20 0 20 400
100
200
300
400
500
600
dE/dx
σK-
σK+
-20 0 20 400
100
200
300
400
500
600
0 10 20 30 400
200
400
600RICH
[nγ]
K+
0 10 20 30 400
200
400
600RICH
[nγ]
K-
Figure 4.2: Histograms of (−2 lnLK)−(−2 lnLπ) from RICH information, number ofphotons nγ used to determine the RICH information, and σK , the number of standarddeviations away from the expected dE/dx for a true kaon, for the positive (1st row)and negative (2nd row) kaons after all other selection criteria have been applied.
49
Figure 4.3: Distribution of (a) mK+K−π0 for |ΔM | < 1 MeV/c2 and (b) ΔM for1.84 < mK+K−π0 < 1.89 GeV/c2 after passing all other selection criteria discussedin the text. The solid curves (red) show the results of fits to the mK+K−π0 and ΔMdistributions, respectively. The background level in each plot is shown by a dashedhorizontal line (black). The vertical lines in (a) and the left-most set of vertical linesin (b) denote the signal region. The right-most set of vertical lines in figure (b) denotethe ΔM sideband used for estimation of the background shape.
50
4.3 Event Candidates
We reconstruct the decay chain D∗+ → π+s D0, D0 → K+K−π0 with the requirement
that the D∗+ momentum be at least as large as one-half of its maximum allowed value
(pD∗max =
√E2
beam
c2− M2
D∗+c2 = 4.88 GeV /c). This suppresses large combinatoric
backgrounds and also removes D∗s from B-meson decays (D∗-mesons which come
from B-mesons at rest cannot have a momentum of more than about 2.26 GeV/c).
The D0 candidate invariant mass mK+K−π0 and invariant D∗+ − D0 mass difference
ΔM ≡ mK+K−π0π+s− mK+K−π0 − (mD∗+ − mD0) are calculated for each candidate,
where mD∗+ and mD0 are the world average values of 2010.0±0.4 MeV and 1864.5±0.4
MeV, respectively, taken from Reference [1]. The distributions of mK+K−π0 and ΔM
are shown in Figure 4.3. We fit each of the distributions to the sum of two bifurcated
Gaussians plus a background shape which is constant (for mK+K−π0) or parabolic
(for ΔM). A sum of two bifurcated Gaussians is defined as the sum of two Gaussian
(Normal) distributions with different widths on either side of the mean (μ). The
mean is the same for both Gaussians. The form used for each Gaussian is:
A
2σA
σA+σB
σA
√2π
e−x−μ
2σ2A for x < μ
A
2σB
σA+σB
σB
√2π
e−x−μ
2σ2B for x >= μ, (4.1)
where A, σA, σB, and μ are parameters in the fit for each Gaussian, x is the variable
representing the distribution we are interested in (mK+K−π0 or ΔM), and the ratio of
both of the widths for the two Gaussians is same (σB2
σB1= σA2
σA1). This shape accounts
for the fact that the signal is slightly broader at lower mass due to uncertainty in
reconstructing the π0 momentum. The parabolic background is a simple fit to a 2nd
order function A(x − a)2 + B(x − a) + C, where a is the x offset, and A, B, and C
are the constants which are fitted.
51
The purpose of these fits is to determine the signal fraction, which is the proportion
of signal events (the events we believe to be true D0s) to background events in the
“signal region” where we expect the vast majority of our signal to be. We select a
signal region defined by 1.84 < mK+K−π0 < 1.89 GeV/c2 and |ΔM | < 1 MeV/c2,
which contains 735 D0 → K+K−π0 candidates. We determine the signal fraction by
integrating the signal fit over the signal region, and dividing by the integral of the
combined signal and background fit over the signal region. We do this separately for
the mK+K−π0 and ΔM plots, finding signal fractions of (79.1±5.3)% and (84.6±6.3)%,
respectively. The average signal fraction is then (81.8 ± 6.3 ± 2.8)%, where the
systematic error is half of the difference between the signal fraction from the fits to
mK+K−π0 and ΔM .
4.4 The Dalitz Plot
We expect CP violation in D decay to be negligible and assume the amplitudes for
D0 → K∗−K+ and D0 → K∗+K− are equal to the amplitudes for charge-conjugated
modes D0 → K∗+K− and D
0 → K∗−K+, respectively. This allows us to double
our statistics in a single Dalitz plot by combining flavor-tagged D0 → K+K−π0 and
D0 → K+K−π0 candidates and choosing the m2
K−π0 variable for one to be the m2K+π0
variable for the other (and vice versa). The inclusion of charge conjugate modes is
implied throughout this chapter.
Figure 4.4(a) shows the Dalitz plot distribution (a scatter plot of m2K+π0 vs.
m2K−π0) for the D0 → K+K−π0 candidates satisfying the requirements described
in Sections 4.2 and 4.3. Charge conjugation is implied, so D0 → K+K−π0 candidates
are on this plot as well, but they are plotted as m2K−π0 vs. m2
K+π0 and superimposed
with the D0 candidates. The density of events is clearly not uniform across the Dalitz
plot, indicating that there is physics which dictates the probability density function.
52
Figure 4.4: (a) The Dalitz plot distribution for D0 → K+K−π0 candidates. (b)-(d)Projections onto the m2
K+π0 , m2K−π0 , and m2
K+K− axes of the results of Fit 1 (discussedin Section 4.10) showing both the fit (curve) and the binned data sample.
53
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
δNR
= 0° δNR
= 90°
δNR
= 180° δNR
= 270°
mK -
π 02 (GeV 2/c 4) m
K+ π
02 (G
eV2 /c
4 )
mK*
2m K*
2 mK -
π 02 (GeV 2/c 4) m
K+ π
02 (G
eV2 /c
4 )
mK*
2
m K*2
mK -
π 02 (GeV 2/c 4) m
K+ π
02 (G
eV2 /c
4 )
mK*
2
m K*2 m
K -π 0
2 (GeV 2/c 4) mK
+ π0
2 (GeV
2 /c4 )
mK*
2
m K*2
Figure 4.5: An example of the effect of a non-resonant (NR) contribution interferingwith two K∗ resonances. The K∗s in this example have the same amplitude but arelative phase with each other such that the two resonances have maximal destructiveinterference with each other, and the amplitude aNR is set to be 5aK∗ . When therelative phase of the NR contribution to the K∗+ phase, δNR, is 0◦ or 180◦ thereis not a noticeable effect on the K∗ lobes. However, a phase δNR = 90◦ makes theK∗− enhanced (depleted) at low (high) K+π0 invariant mass squared, and the K∗+
enhanced (depleted) at high (low) K−π0 invariant mass squared. A phase δNR = 270◦
makes the K∗+ enhanced (depleted) at low (high) K−π0 invariant mass squared, andthe K∗− enhanced (depleted) at high (low) K+π0 invariant mass squared.
54
The enhanced bands perpendicular to the m2K−π0 and m2
K+π0 axes at an invariant
mass-squared of m2Kπ ≈ 0.8 GeV2/c4 correspond to K∗(892)− and K∗(892)+ reso-
nances, respectively. The φ(1020) can be seen as a diagonal band along the upper
right edge of the plot. Recall the reason why the φ resonance appears this way on
the m2K+π0 vs. m2
K−π0 plot: Due to energy and momentum conservation, m2K+K−
is a function of m2K+π0 and m2
K−π0 . The vector (spin 1) nature of these resonances
is evident from the depleted region in the middle of each band, which comes from
the cos(θ) dependence discussed in Section 2.3. The nearly missing bottom lobe of
the K∗(892)− band and the enhanced left lobe of the K∗(892)+ band hint that these
resonances are interfering with opposite phases with an S-wave amplitude (such as
the non-resonant contribution) under these resonances. An example of how the non-
resonant contribution can affect the K∗ lobes is shown in Figure 4.5.
The major resonances can be more easily be seen in the projections of the Dalitz
plots (Figure 4.4(b)-(d)), as peaks in the two-particle invariant masses. There are
multiple peaks in the projection plots, but some are simply reflections of other reso-
nances. To visualize both peaks that are resonances in a certain projection and peaks
that are reflections, we can arrange the projections as in Figure 4.6.
4.5 The Matrix Element
In most Dalitz plot analyses, choosing the form of the matrix element M is the most
difficult part of the whole analysis procedure. The isobar model, a coherent sum of
Breit-Wigner amplitudes, is the simplest approach, but is known to only be reliable
in cases where the resonances being modeled are fairly narrow, far from threshold,
and well separated. This should be a reasonable approximation for the dominant
K∗(892) and φ contributions in this analysis, but perhaps not as reasonable for other
resonant components. For the purposes of this analysis the isobar model is sufficient,
55
K*+
K*-
φ
Figure 4.6: Dalitz plot projections properly rotated to show how they are obtainedfrom the Dalitz plot. The thicker straight lines correspond to resonances which decayto the two particles in the respective invariant mass-squared projection, and thethinner straight lines correspond to reflections which do not decay to the two particlesin the respective invariant mass-squared projection.
56
Resonance r mr (GeV/c2) Γr (GeV/c2) SpinK∗(892)± 0.8917 0.0508 1φ(1020) 1.0190 0.0043 1
non-resonant uniform uniforma0(980) 0.9910 0.0690 0f0(1370) 1.3500 0.2650 0
K∗0(1430)± 1.4120 0.2940 0
K∗2(1430)± 1.4260 0.0985 2f0(1500) 1.5070 0.1090 0f ′
2(1525) 1.5250 0.0730 2κ± 0.8780 0.4990 0
Table 4.2: The masses and widths of resonances r considered in this analysis [1,45–48].
but other models do exist and have been tested in this analysis (see Appendix A).
We express the complex amplitude for a D0 decay to the jth quasi-two-body state
as ajeiδjB(k)
j , where aj is real and positive and B(k)j is the Breit-Wigner amplitude for
resonance j with spin k of Equation 2.3 described in Chapter 2. The amplitudes B(k)j
have no particular normalization.
We consider thirteen resonant components (see Table 4.2) as well as a uniform
non-resonant contribution described by the simple complex amplitude aNReiδNR . The
matrix element M is most generally written as a sum over the individual resonant
amplitudes and the non-resonant contribution:
M =∑
j
ajeiδjB(k)
j + aNReiδNR . (4.2)
4.6 Method of Maximum Likelihood
Once we have a set of D0 → K+K−π0 candidates, we want to match these data with
a given theoretical model which constitutes M. The most useful general method
of choosing the free parameters in a given model of M is the method of maximum
likelihood [1].
57
The basic idea of the method of maximum likelihood is that the best model is
consistent with the observation. We begin by assuming that our event distribution
came from the probability distribution function, L, whose form is what we expect from
theory, and L has parameters whose values are unknown. The method of maximum
likelihood forms a likelihood which is the product over all of our events of the value
of L for each event. By finding an optimal set of parameters, we seek to maximize
this likelihood. This effectively determines the parameters for which the data is most
likely.
We can explain why the method of maximum likelihood works. If the observed
events do not match where our model expects the events to be, L at those events
will be very low, and L at each of the events will multiply to a small number. If
the observed events are where our model expects them to be, then for each point
the probability distribution function will be high, and L at each of the events will
multiply to a large number. In order to find the best model based on the physics
which we believe dictates the probability distribution function, we adjust the free
parameters until we get the highest value of∏
events
L.
For our analysis, we can form a probability distribution function which has as
its parameters the amplitudes and phases of the various resonances and non-resonant
contribution described in Section 4.5, and the signal fraction f as described in Section
4.3:
L = fE(y, z)|M|2∫
E(y, z)|M|2 dDP + (1 − f )B(y , z )∫
B(y , z ) dDP , (4.3)
where M is the matrix element “model” for the decay, E is a polynomial representing
the efficiency shape, and B is a polynomial representing the background shape. E
and B are described in Section 4.7. M, E, and B are all functions of the two Dalitz
plot variables chosen for fitting. (1 − f) is the fraction of event candidates that are
58
believed to be background. In the probability distribution function L, the efficiency-
corrected matrix element squared and the background shape are each normalized over
the Dalitz plot “area” (DP) defined by the kinematic limits of the D0 decay to the two
Dalitz plot variables. These normalizations ensure that L represents what we believe
to be probability distribution function; the integral of the probability distribution
function over DP is 1, as expected.
The likelihood,∏
events
L, tells us how likely it is that our data on the Dalitz plot fit
into a model with a particular set of parameters. We use a MINUIT [49] based fitter
to minimize the negative logarithmic likelihood function L:
L =∑
events
−2 ln L, (4.4)
which is effectively the same as maximizing∏
events
L. The efficiency and background
shapes are determined as described in Section 4.7. We allow our signal fraction f to
float in the fits, unless otherwise noted.
Note that one can multiply all of the pieces of the matrix element by a complex
constant (i.e. an amplitude and phase) and the probability distribution function
would be exactly the same. Thus, we can only fit for the shape of |M|2, not its
absolute value.
Since Dalitz plot analyses are only sensitive to relative phases and amplitudes, we
may arbitrarily define the amplitude and phase for one of the two-body decay modes.
The mode with the largest rate, K∗+K−, is assigned an amplitude aK∗+K− = 1
and phase δK∗+K− = 0◦. Measuring the remaining amplitudes and relative phases
of the resonances, which will be free parameters in our fit, thus gives us a detailed
understanding of the substructure of this decay.
At this point, we need to choose the invariant-mass squared variables we will use
to parameterize our Dalitz plot. We choose to fit the Dalitz plot in (m2K+π0 , m2
K+K−)
59
rather than (m2K−π0 , m2
K+π0) or (m2K+K− , m2
K−π0). m2K+K− is chosen because the
momentum resolution of tracks is better than the momentum resolution of π0s. m2K+π0
is chosen because the resonance with the largest rate, K∗+, decays to K+π0.
4.7 Efficiency and Background
4.7.1 Efficiency
The kinematically-allowed range of invariant mass-squared values define all of the
possible states (or “phase space”) of the system. The efficiency of our ability to
reconstruct a D0 → K+K−π0 event is not constant over the entire phase space.
Since a fit to the Dalitz plot will yield only relative values for amplitudes, only the
shape of the efficiency across the Dalitz plot is important, not its absolute value. We
study efficiency by generating a large Monte Carlo sample of D0 → K+K−π0 signal
events, which are equally likely to decay at any point in the allowed phase space, and
running these events through the same analysis code that is used on data. The shape
of the remaining event distribution is a direct measure of the efficiency.
A 2-D cubic polynomial is used to fit the shape of the efficiency. We only need to
fit to two of the three possible invariant mass combinations x ≡ m2K−π0 , y ≡ m2
K+π0 ,
and z ≡ m2K+K− , and choose to fit to y and z. The form of the efficiency function
E(y, z) is given by:
E(y, z) = E0 + Eyy + Ezz
+ Eyyy2 + Eyzyz + Ezzz
2
+ Eyyyy3 + Eyyzy
2z + Eyzzyz2 + Ezzzz3. (4.5)
Table 4.3 shows the best-fit parameters for the efficiency shape. Figure 4.7 shows
the distribution of events used to determine the efficiency across the Dalitz plot as
60
E0 Ey Ez Eyy Eyz
1 (fixed) -3598 ± 1027 977.4 ± 488.4 3459 ± 1053 2686 ± 995.3
Ezz Eyyy Eyyz Eyzz Ezzz
-244.7 ± 298.7 -755.3 ± 283.2 -1184 ± 372.0 -423.5 ± 217.2 -18.85 ± 65.54
Table 4.3: Best-fit parameters for the efficiency shape.
well as the result of our fit to these events.
4.7.2 Background
Figure 4.3 shows that background D0 → K+K−π0 candidates are clearly present even
after all selection requirements are imposed. Since there is no way to sort individual
events into “signal” and “background” categories, we must try to understand the
fraction of these events and their collective shape on the Dalitz plot since this will
allow us to take backgrounds into account when extracting physics information from
our Dalitz plot fit.
Our approach is to select events in a ΔM sideband region (the “Background
Region”), which are known to be almost entirely background, and use them to model
the shape of background events in the signal region. Figure 4.8(a) is a scatter plot
of the D∗+ − D0 mass difference versus D0 mass for D0 → K+K−π0 candidates
satisfying all other selection requirements. The lower rectangular region represents
the “Signal Region” and hence the events that form the data sample we use for
our Dalitz plot analysis. In other words, each point inside the lower rectangle of
Figure 4.8(a) corresponds to a point on the Dalitz plot shown in Figure 4.4(a). The
top rectangular region in Figure 4.8(a) contains very little signal and is used to model
the shape of the background across the Dalitz plot and Figure 4.8(b) is a Dalitz plot of
the events in this region. The background events are fitted to extract the background
shape parameters. The signal and background regions have the same extent along
the D0 mass axis, and the extent of the background sideband along the ΔM axis is
61
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
0.7
1.1
1.5
1.9
0.7
1.1
1.5
1.9
mK+π0
2 (GeV
2/c
4)
mK-π
02 (GeV
2 /c4 )(a
)
(d)
(b)
(c)
Events per 0.03 GeV2/c
4
Fig
ure
4.7:
(a)
Sca
tter
plo
tof
Mon
teC
arlo
even
tsuse
dto
study
the
effici
ency
acro
ssth
eD
alit
zplo
tan
d(b
-d)
pro
ject
ions
ofth
ese
even
tsal
ong
wit
hth
ere
sult
ofa
fit
usi
ng
the
effici
ency
funct
ion.
62
0.7 1.1 1.5 1.9
0.7
1.1
1.5
1.9
1.7 1.8 1.9 2.0-10
-5
0
5
10
15
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
mK+K-
π0 (GeV/c2)
ΔM
(MeV
/c2 )
(a) (b)
mD0
Figure 4.8: (a) Scatter plot of ΔM versus mK+K−π0 for event candidates satisfyingall other requirements. The black boxed area (the bottom box) is the Signal Regionwhich contains our signal candidates, and the red boxed area (the top box) is theBackground Region which contains events which we use to estimate the shape of ourbackground. (b) a Dalitz plot of the events from the Background Region of plot (a).
defined to be 0.003 GeV /c2 < ΔM < 0.010 GeV /c2.
Note that this technique assumes that the background events in both the ΔM
signal and ΔM sideband regions have the same distribution across the Dalitz plot.
This is a good assumption since ΔM depends predominantly on our measurement of
the slow pion from the D∗ decay and only weakly on the D0 decay itself.
In order to model the background, we use a 2-D cubic polynomial plus terms
to represent particles which we expect to be in the background. The form of the
63
background function B(y, z) is given below:
B(y, z) = B0 + Byy + Bzz
+ Byyy2 + Byzyz + Bzzz
2
+ Byyyy3 + Byyzy
2z + Byzzyz2 + Bzzzz3
+ BK∗+ |A(K∗+)|2
+ Bφ|A(φ)|2
+ BK∗−|A(K∗−)|2. (4.6)
The last three terms in B represent K∗+, K∗−, and φ particles which are present in
the background but are not from D0 → K+K−π0 decays. Note that these amplitudes
do not interfere.
B0 By Bz Byy Byz
1 (fixed) 20.58 ± 64.32 11.80 ± 35.89 92.65 ± 60.44 -89.03 ± 55.36
Bzz Byyy Byyz Byzz Bzzz
9.596 ± 25.08 -43.19 ± 23.56 18.84 ± 17.76 8.572 ± 11.94 -1.482 ± 4.688
BK∗+ Bφ BK∗−
.02742 ± .02200 .03254 ± .02597 .01933 ± .01939
Table 4.4: Best-fit parameters for the background shape.
Table 4.4 shows the best-fit parameters for the background shape. Figure 4.9
shows the projected background events from the ΔM sideband and the result of
fitting the background Dalitz plot with the background function.
4.8 Overview of Fitting Technique
In the case of our D0 → K+K−π0 analysis we can guess what the main quasi two-body
contributions are just by examining the Dalitz plot in Figure 4.4(a). We can see clear
enhancements that correspond to D0 → K+K∗−, D0 → K−K∗+ and D0 → φπ0. Any
64
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.9:
The
pro
ject
edbac
kgr
ound
even
tsfr
omth
eΔ
Msi
deb
and
and
the
resu
ltof
fitt
ing
the
bac
kgr
ound
Dal
itz
plo
tto
a2-
Dcu
bic
pol
ynom
ialplu
snon
-inte
rfer
ing
K∗
and
φco
ntr
ibuti
ons
(lin
e).
65
matrix element we hypothesize for the D0 → K+K−π0 decay must include these three
resonances. However, there may be additional broad resonances present which are
not clearly visible in Figure 4.4(a). It is precisely this, the inclusion and parametriza-
tion of broad components that are difficult to distinguish from non-resonant decays,
efficiency variations, and/or backgrounds, that is the main challenge in most Dalitz
analyses.
The main goal of this analysis is to measure the strong phase difference δD and
relative amplitude rD. In terms of the amplitudes and phases in the matrix element,
δD and rD are defined by the following:
rDeiδD =aK∗−K+
aK∗+K−ei(δK∗−K+−δK∗+K− ), (4.7)
where rD in Equation 4.7 is defined as real and positive. The strong phase difference
is equivalent to the overall phase difference for these decays due to our assumption
that CP violation in D decays is negligible. δD is the difference in phase between the
D0 → K+K∗− and D0 → K−K∗+ components on this Dalitz plot, and can therefore
be extracted directly from a fit to these data.
Measuring δD and rD is in principle very simple, however in practice it is compli-
cated by a couple of things. One issue is that the width of the φ is comparable to
our resolution. thus in order to model the φ properly we considered the possibility of
convoluting this resonance with our invariant mass squared resolution in our model.
However, we believe that the width of the φ is close enough to our resolution so that
we may use the documented width of the φ in our analysis and choose to evaluate the
systematic uncertainty on rD and δD as a result of this choice. Also, the uncertainty
in modeling possible broad resonant contributions, especially resonant contributions
in regions of the Dalitz plot which affect δD and rD, may affect our results. We will
approach this problem by adding various additional resonant components which seem
66
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
δD
= 0° δD
= 180°
mK -
π 02 (GeV 2/c 4) m
K+ π
02 (G
eV2 /c
4 )
mK*
2
m K*2 m
K -π 0
2 (GeV 2/c 4) mK
+ π0
2 (GeV
2 /c4 )
mK*
2
m K*2
0.5
1
1.5 0.5
1
1.50
500
1000
0.5
1
1.5
Figure 4.10: Interfering K∗+ and K∗− resonances on the Dalitz plot. The way wehave defined our amplitudes results in a phase convention where maximal destructiveinterference between the K∗s occurs at δD = 0◦. Maximal constructive interferenceoccurs when δD = 180◦. Note that the effect of the interference is most noticeable inthe region where the K∗ resonances overlap.
reasonable and investigating their effect on rD and δD.
It is important to note that our sign convention for our amplitudes implies that
δD ≡ δK∗−K+ − δK∗+K− = 0◦ indicates maximal destructive interference between the
K∗ amplitudes (see Figure 4.10). This is due to our choice of defining amplitudes
which decay to different resonant two-body modes by simple cyclic permutations, as
discussed in Section 2.2.
The unbinned maximum likelihood fitter used was originally developed by Tim
Bergfeld for his study of D0 → K−π+π0 [14] and has subsequently been modified
by many others doing CLEO Dalitz plot analyses. Recent modifications include the
capability to fit only parts of the Dalitz plot (see Subsection 4.10.5) and the ability
to implement the K-matrix formalism (see Appendix A).
The fitter excludes events which are not within the kinematic limits of the D0 →K+K−π0 decay. 24 of the 735 events on the Dalitz plot fall outside these limits in
the (m2K+π0 , m2
K+K−) plane and are excluded, leaving 709 events which are fitted. An
event may be excluded because it is background (the tracks and showers reconstruct to
a particle with the D0 mass, but it is not a D0 → K+K−π0 decay because conservation
67
of momentum would not allow it) or because it is signal and the candidate falls just
outside the kinematic boundary due to limits in how well we can measure the decay.
The efficiency and background functions, E and B, respectively, are those de-
scribed in Section 4.7, the signal fraction f (see Equation 4.4) is allowed to float in
the fit, and the fits were done in the (m2K+π0 , m2
K+K−) plane, unless otherwise stated.
In our fitter, for resonances that decay to the third two-particle combination, K− and
π0, m2K−π0 is calculated from the two mass-squared variables that are used in the fit.
In order to get some idea of how much of each type of resonant three-body decay
is in each particular fit, the fit fraction (FF ) for the resonant component j of a given
fit may be calculated as follows:
FFj =
∫ ∣∣ajeiθjBj
∣∣2dDP∫ |M|2dDP , (4.8)
where we may substitute “1” for “Bj” when calculating the fit fraction for the non-
resonant contribution.
We don’t get the exact fraction of how much of each resonance is in the fit,
since the amplitude squared in the numerator of Equation 4.8 only tells us what we
would expect if that amplitude was not interfering with other resonant amplitudes.
Therefore, the sum over all the fit fractions j, known as the total fit fraction, does
not necessarily have to add up to 100%. A total fit fraction of greater than 100%
implies overall destructive interference (the denominator of Equation 4.8 is small),
while a total fit fraction of less than 100% implies overall constructive interference
(the denominator of Equation 4.8 is large).
68
4.9 Goodness of Fit
Judging the quality of a Dalitz plot fit can be difficult. We may make some qualitative
judgements by simply looking at our fits, although the best way to do this isn’t
obvious. There are two quantitative methods that we use. The first is a very natural
extension of the fitting method. Since we are minimizing the summed log likelihood
(see Equation 4.4), we can simply look at the negative logarithmic likelihood function
(L) and see how it differs between fits. L does not in itself give us a very intuitive
indication of the quality of the fit. However, fits which have a significantly higher
value of L for the same number of events are considered to be inferior. That said,
it is important to be wary of fits which have a significantly lower value of L because
our model has too many parameters.
Another measure of goodness-of-fit is the Significance Level (S.L.) which is calcu-
lated for each fit which used the entire Dalitz plot:
S. L. =1
2
(1 − erf
(L − < L >
σ<L>
√2
)), (4.9)
where L is the negative logarithmic likelihood function, < L > is the expected mean
for the negative logarithmic likelihood function assuming the events are truly dis-
tributed according to the likelihood function which gives the best fit, and σ<L> is
the standard deviation associated with < L >. The procedure is detailed in Refer-
ences [14, 50]. It is important to note that the S.L. only gives a measurement of the
goodness of fit assuming the fit function correctly describes the distribution.
There are other considerations which can be very useful in determining the via-
bility of a fit. If there are too many resonances in a fit, it is likely that the fit fraction
of some of those resonances will be insignificant. If a particular fit has insignificant
resonances, it is highly likely that the resonance content of the fit does not represent
what is in our data (at least to the precision which we can measure). Also, if the sig-
69
nal fraction that the fit prefers differs substantially from the measured signal fraction
(as measured in Section 4.3), it is likely that the fit is not accurate.
4.10 Fit Results
The strategy we used when fitting the data was to start with the minimum sensible
number of Dalitz plot components (in our case: K∗+, K∗−, φ, and a “non-resonant”
contribution, all of which are allowed to interfere with each other) and then observe
the effects of excluding the non-resonant part and of including additional resonances
to the fit. Table 4.2 gives a list of the resonances that were tried. The first three rows
represent the components of the “nominal” fit and the rest are various additional
resonances.
The fit projections for each fit are shown in Figures 4.11 through 4.32, 4.34 through
4.36, and 4.38 through 4.45, and a descriptive summary of all fits is given in Table 4.5.
The first thirteen fits and the last fit were done using data across the entire Dalitz
plot, and the other fits were partial fits that excluded specific regions of the Dalitz
plot (reasons for which are discussed in Subsection 4.10.5).
4.10.1 Nominal Fit
Figure 4.11 (Fit 1) shows the results of the simplest possible reasonable fit to the
data, the nominal fit. In this fit the matrix element M is composed of the three
obvious resonances seen in Figure 4.4(a) (K∗+, K∗−, and φ) plus a uniform non-
resonant (NR) component. Notice that these provide a very good fit to the data. A
measurement of δD = 332◦ (or equivalently, −28◦) indicates destructive interference
between the K∗s.
Three-body decays of D mesons are expected to be dominated by resonant two-
body decays [35–39], so it is worthwhile to see what the fit looks like without having
70
Fit
Res
onan
ces
use
dR
egio
nL
S.L
.f
(%)
r Dδ D
φK
∗±N
Rκ±
K∗± 0
a0
f 0K
∗± 2f 2
1x
xx
Who
leD
P−9
7.4
18.6
%85
.9±
5.6
0.52
±0.
0533
2◦±
8◦
2x
xW
hole
DP
−27.
015
.4%
56.8
±3.
40.
60±
0.06
171◦
±28
◦
3x
xx
Who
leD
P−1
10.9
17.2
%80
.1±
4.7
0.52
±0.
0531
3◦±
9◦
3ax
xx
xW
hole
DP
−117
.919
.9%
86.2
±5.
50.
50±
0.06
318◦
±9◦
4x
xx
xW
hole
DP
−122
.322
.8%
86.3
±5.
40.
43±
0.07
327◦
±12
◦
5x
xx
xW
hole
DP
−112
.627
.5%
84.9
±5.
30.
46±
0.07
314◦
±12
◦
6x
xx
Who
leD
P−8
0.6
28.0
%78
.9±
5.4
0.35
±0.
0514
◦±
16◦
7x
xx
1370
Who
leD
P−6
7.6
17.5
%65
.4±
4.4
0.53
±0.
0624
2◦±
19◦
8x
xx
x1370
Who
leD
P−1
03.9
26.0
%82
.2±
5.2
0.35
±0.
088◦
±27
◦
9x
xx
x1370
xW
hole
DP
−113
.429
.1%
82.5
±4.
40.
30±
0.06
45◦±
34◦
10x
xx
xx
1370
xW
hole
DP
−146
.636
.0%
86.3
±5.
70.
44±
0.09
353◦
±18
◦
11x
xx
1500
Who
leD
P−1
01.4
17.7
%86
.1±
5.9
0.55
±0.
0532
4◦±
10◦
12x
xx
xW
hole
DP
−98.
421
.5%
84.7
±6.
10.
53±
0.05
332◦
±8◦
1xx
xK
∗ba
nds
825.
5N
/A87
.7±
5.6
0.50
±0.
0533
3◦±
14◦
2xx
K∗
band
s88
5.1
N/A
61.3
±6.
30.
68±
0.07
290◦
±27
◦
3xx
xK
∗ba
nds
829.
1N
/A84
.6±
6.1
0.57
±0.
0630
4◦±
17◦
1nx
xN
oφ
1.05
331.
2N
/A80
.5±
6.2
0.52
±0.
0533
7◦±
9◦
1np
xx
xN
oφ
1.05
321.
6N
/A84
.1±
6.3
0.50
±0.
0533
4◦±
9◦
1na
xx
No
φ1.
1522
0.6
N/A
84.9
±5.
90.
50±
0.05
343◦
±9◦
1nap
xx
xN
oφ
1.15
215.
8N
/A84
.9±
6.0
0.48
±0.
0533
8◦±
11◦
1nb
xx
No
φ1.
2612
3.5
N/A
83.2
±6.
10.
50±
0.05
338◦
±10
◦
1nbp
xx
xN
oφ
1.26
122.
1N
/A83
.1±
6.4
0.50
±0.
0633
1◦±
12◦
1sx
x“L
”ar
ea33
1.2
N/A
82.1
±6.
20.
50±
0.05
334◦
±10
◦
3fx
xx
Who
leD
P−1
18.3
16.2
%79
.7±
5.3
0.53
±0.
0530
6◦±
11◦
Tab
le4.
5:Sum
mar
yof
fits
.See
Tab
le4.
2fo
rm
ore
info
rmat
ion
abou
tea
chre
sonan
ce,an
dse
eth
efo
llow
ing
figu
res
for
det
ails
.
71
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.11
:D
alit
zplo
tpro
ject
ions
for
Fit
1sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
72
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.12
:D
alit
zplo
tpro
ject
ions
for
Fit
2sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
73
a non-resonant contribution. Figure 4.12 (Fit 2) shows what happens when we simply
remove the non-resonant (NR) component from Fit 1. While the fit may look visually
acceptable, the fit wants the signal fraction, f , to be significantly smaller (≈ 57%)
than the value measured in Section 4.3 (≈ 82%). L, the negative logarithmic likeli-
hood function, is about 70 higher than for the nominal fit.
4.10.2 κ Fit
Removing the non-resonant contribution had a significant negative effect on the nom-
inal fit, hence we expect there to be a large, broad S-wave component in the region
of low Kπ mass in order to create the type of interference observed on our Dalitz
plot. Having a non-resonant contribution certainly provides this S-wave component,
but it is also possible to have a large low Kπ mass S-wave if there is a broad scalar
resonance which decays to Kπ. While such a resonance has not been published in
the Review of Particle Physics [1], there is some evidence for a low-mass broad scalar
resonance which decays to Kπ called the κ [45]. Since we expect three-body decays of
D mesons to be dominated by resonant two-body decays, the κ is worth considering
in our fit as an alternative to the non-resonant contribution.
Figure 4.13 (Fit 3) shows what happens when we replace the non-resonant (NR)
component with a κ+ and a κ−. The fit remains good and rD is unaffected, but the
phase difference δD changes from the nominal fit, although the change is not drastic.
The signal fraction from the fit is once again consistent with our measured signal
fraction.
4.10.3 Results of Nominal Fit and κ Fit
Table 4.6 contains the results for our nominal fit with statistical errors, including
calculated fit fractions. The signal fraction from the fit is (86 ± 6)%. Since the K∗+
74
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.13
:D
alit
zplo
tpro
ject
ions
for
Fit
3sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
75
Resonance Amplitude Phase Fit FractionK∗+ 1.00 (fixed) 0◦ (fixed) (46.1 ± 3.1)%K∗− 0.52 ± 0.05 332◦ ± 8◦ (12.3 ± 2.2)%φ 0.64 ± 0.04 326◦ ± 9◦ (14.9 ± 1.6)%
Non-Resonant 5.62 ± 0.45 220◦ ± 5◦ (36.0 ± 3.7)%Total Fit Fraction: (109.3 ± 5.5)%
Table 4.6: Amplitude, phase, and fit fraction results for our nominal fit (Fit 1) withstatistical errors.
amplitude and phase were fixed, we require an independent method to determine the
statistical errors on the fit fractions. We use the full covariance matrix from the fitter
and final parameter values to generate 5000 sample parameter sets. For each set, the
fit fractions are calculated and recorded in histograms. These histograms are fit with
a Gaussian to extract their width, which is used as the statistical error on the fit
fraction.
The signal and background in the Dalitz plot projections of the nominal fit (Fit
1) is shown in Figure 4.14. A graphical version of how much of each resonance is in
the total signal for Fit 1 is shown in the projections in Figures 4.15, 4.16, and 4.17.
Note that the amounts of each individual resonance are those from the numerator
of the fit fraction in Equation 4.8, and so the sum of the areas of each resonance in
Figures 4.15, 4.16, and 4.17 do not add up to the area of the signal. The total fit
fraction of more than 100% for Fit 1 implies overall destructive interference.
Fit 3, in which we replaced the non-resonant amplitude with a κ+ and a κ−
amplitude, is also a reasonable match to the data. While we are not certain with
what seriousness the κ resonances should be considered, we present results for the fit
with the κ for the purpose of review.
Table 4.7 contains the results for Fit 3 with statistical errors, including calculated
fit fractions. We again generate sample parameter sets to determine the statistical
errors on the fit fractions. The signal fraction from the fit is (80 ± 5)%.
76
01
23
0122436
48
60
72
01
23
020
40
60
80
10
0
12
0
01
23
0612182430
36
Events per 0.03 GeV2/c
4
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Tota
lSi
gn
alB
ackg
rou
nd
Fig
ure
4.14
:T
he
sign
alan
dbac
kgr
ound
inth
eD
alit
zplo
tpro
ject
ions
ofFit
1.
77
0 1 2 3
Even
ts p
er 0
.03
GeV
2 /c4
mK+
π02 (GeV2/c4)
0
12
24
36
48
60
72
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�NR
|2
Figure 4.15: Signal portion of Fit 1 and the component resonance amplitudes squared(m2
K+π0 projection).
78
0 1 2 30
20
40
60
80
100
120
Even
ts p
er 0
.03
GeV
2 /c4
mK+K-
2 (GeV2/c4)
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�NR
|2
Figure 4.16: Signal portion of Fit 1 and the component resonance amplitudes squared(m2
K+K− projection).
79
0 1 2 30
6
12
18
24
30
36
Even
ts p
er 0
.03
GeV
2 /c4
mK-
π02 (GeV2/c4)
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�NR
|2
Figure 4.17: Signal portion of Fit 1 and the component resonance amplitudes squared(m2
K−π0 projection).
80
01
23
0122436
48
60
72
01
23
020
40
60
80
10
0
12
0
01
23
0612182430
36
Events per 0.03 GeV2/c
4
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Tota
lSi
gn
alB
ackg
rou
nd
Fig
ure
4.18
:T
he
sign
alan
dbac
kgr
ound
inth
eD
alit
zplo
tpro
ject
ions
ofFit
3.
81
0 1 2 30
12
24
36
48
60
72
Even
ts p
er 0
.03
GeV
2 /c4
mK+
π02 (GeV2/c4)
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�κ
+|2
|�κ-|
2
Figure 4.19: Signal portion of Fit 3 and the component resonance amplitudes squared(m2
K+π0 projection).
82
0 1 2 30
20
40
60
80
100
120
Even
ts p
er 0
.03
GeV
2 /c4
mK+K-
2 (GeV2/c4)
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�κ
+|2
|�κ-|
2
Figure 4.20: Signal portion of Fit 3 and the component resonance amplitudes squared(m2
K+K− projection).
83
0 1 2 30
7.5
15
22.5
30
Even
ts p
er 0
.03
GeV
2 /c4
mK-
π02 (GeV2/c4)
Signal|�
φ|2
|�K*+|
2
|�K*-
|2
|�κ
+|2
|�κ-|
2
Figure 4.21: Signal portion of Fit 3 and the component resonance amplitudes squared(m2
K−π0 projection).
84
Resonance Amplitude Phase Fit FractionK∗+ 1.00 (fixed) 0◦ (fixed) (48.1 ± 4.5)%K∗− 0.52 ± 0.05 313◦ ± 9◦ (12.9 ± 2.6)%φ 0.65 ± 0.05 334◦ ± 12◦ (16.1 ± 1.9)%κ+ 1.78 ± 0.43 109◦ ± 17◦ (12.6 ± 5.8)%κ− 1.60 ± 0.29 128◦ ± 17◦ (11.1 ± 4.7)%
Total Fit Fraction: (99.0 ± 9.3)%
Table 4.7: Amplitude, phase, and fit fraction results for Fit 3 (with κ± instead of anon-resonant contribution) with statistical errors.
The signal and background in the Dalitz plot projections of Fit 3 is shown in
Figure 4.18. A graphical version of how much of each resonance is in the total signal
for Fit 3 is shown in the projections in Figures 4.19, 4.20, and 4.21. Note that the
amounts of each individual resonance are those from the numerator of the fit fraction
in Equation 4.8, and so the sum of the areas of each resonance in Figures 4.19, 4.20,
and 4.21 do not add up to the area of the signal. For Fit 3 the total fit fraction is not
statistically conclusive as overall destructive or constructive interference. However,
a phase of 313◦ ± 9◦ (alternatively, −47◦ ± 9◦) does imply destructive interference
between the K∗ resonances.
To compare the fits, we can overlay them as shown in Figure 4.22. They look very
similar but do have very subtle differences.
4.10.4 Other Fits to the Entire Dalitz Plot
Figure 4.23 (Fit 3a) shows what happens when we add the non-resonant (NR) compo-
nent to the set of resonances in Fit 3. The fit remains good and the phase difference
δD and relative amplitude rD change slightly from Fit 3. The κ+ does not have a
significant fit fraction, though (FFκ+ = (1.6 ± 1.8)%).
Figure 4.24 (Fit 4) shows what happens when we replace the non-resonant (NR)
component with four rather broad resonances (κ± and K∗0(1430)±). The fit improves
85
mK
+K-2 (G
eV2 /c
4 )m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fit
1Fi
t 3
Fig
ure
4.22
:Fit
1pro
ject
ions
(blu
e,das
hed
line)
over
laye
don
top
ofFit
3pro
ject
ions
(gre
en,so
lid
line)
.
86
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.23
:D
alit
zplo
tpro
ject
ions
for
Fit
3ash
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
87
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.24
:D
alit
zplo
tpro
ject
ions
for
Fit
4sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
88
slightly and the phase difference δD remains fairly unchanged from Fit 1, while rD
decreases. The K∗0 ’s do not seem to have significant fit fractions, though (FFK∗+
0=
(8.9 ± 7.6)%, FFK∗−0
= (0.9 ± 2.2)%).
Figure 4.25 (Fit 5) shows what happens when we include both a non-resonant
(NR) component and the K∗0(1430)±. In this case we clearly have redundancy in the
basis set since the fit fractions come out completely nonsensical: the NR contribution
is (FFNR = (127±51)%) and the K∗0 ’s are barely significant (FFK∗+
0= (15.0±13.2)%,
FFK∗−0
= (19.0 ± 11.6)%). Indeed, the total fit fraction for all resonances is ≈ 224%
which implies destructive interference on an unrealistic scale.
Figure 4.26 (Fit 6) shows what happens when we replace the non-resonant (NR)
component with K∗0(1430)±. In this case the K∗
0(1430)− resonance does not have a
significant fit fraction (FFK∗−0
= (1.3±2.0)%). This is not hard to understand: We are
attempting to replace a broad non-resonant background with high-mass resonances
whose widths are not big enough to have much influence on the lower mass regions
of the Dalitz plot. Since this basis set is insufficient to cover the Dalitz plot, the fit
in this case is pulled in an unnatural direction.
The same argument holds for Figure 4.27 (Fit 7). The replacement of the NR
component with a0(980) (FFa0 = (2.9± 1.6)%) and f0(1370) is not enough to satisfy
the need for a broad low mass feature that a simple non-resonant component can
satisfy very well. Also, the signal fraction for this fit is much lower than the measured
signal fraction.
Figures 4.28 and 4.29 (Fits 8 and 9) show the effects of including all of the ad-
ditional resonances tried in Fits 6 and 7 (i.e. K∗0(1430), a0 and f0), with the extra
addition of the K∗2(1430) in Fit 9. The summed minus log likelihood improves, but
there are many non-significant components.
Figure 4.30 (Fit 10) shows the result of using the same resonances used in Fit 9
with the addition of the κ resonances. In this case we clearly have redundancy in the
89
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.25
:D
alit
zplo
tpro
ject
ions
for
Fit
5sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
90
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.26
:D
alit
zplo
tpro
ject
ions
for
Fit
6sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
91
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.27
:D
alit
zplo
tpro
ject
ions
for
Fit
7sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
92
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.28
:D
alit
zplo
tpro
ject
ions
for
Fit
8sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
93
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.29
:D
alit
zplo
tpro
ject
ions
for
Fit
9sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
94
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.30
:D
alit
zplo
tpro
ject
ions
for
Fit
10sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
95
basis set due to a number of insignificant resonances.
Figure 4.31 (Fit 11) shows a fit that uses the resonances in Fit 1 along with the
f0(1500). This fit did not converge in the fitter at first with the default starting
values (aj = 1 and δj = 0 for all resonances j). When I changed the starting values
for amplitudes and phases of the fit to those in Fit 1, the fitter did converge. rD is
consistent with Fit 1 and δD is lower, but the f0(1500) is not significant in this fit
(FFf0(1500) = (1.0 ± 1.1)%).
Figure 4.32 (Fit 12) is a fit that uses the resonances in Fit 1 along with the
f ′2(1525). δD and rD remain fairly unchanged from Fit 1, but the fit fraction for the
f ′2(1525) is not significant (FFf ′
2(1525) = (0.3 ± 0.6)%).
It is useful to summarize what we have learned from Fits 1-12. First, the Dalitz
plot is well fitted by a simple combination of K∗+, K∗−, φ, and a uniform non-resonant
contribution (Fit 1). The non-resonant contribution to Fit 1 is significant; its fit frac-
tion is large (≈ 36%) and removing it makes the fit significantly less accurate (Fit 2).
The resonance that provides the best progress in replacing the non-resonant contri-
bution is the scalar κ; however, the existence of the κ is unconfirmed by Reference [1].
The addition of other sub-modes can improve the appearance of the fit but are not
well motivated since these tend to have masses and widths which do not make them
ideally suited for providing the broad low-mass feature that the simple non-resonant
component can provide. When too many resonances which decay to the same par-
ticles (for example, many resonances which decay to K+K−) are used the fitter has
problems, as evidenced by fit fractions which indicate destructive interference at lev-
els which do not make much sense or multiple resonances which may be insignificant
in the fit.
96
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.31
:D
alit
zplo
tpro
ject
ions
for
Fit
11sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
97
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.32
:D
alit
zplo
tpro
ject
ions
for
Fit
12sh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
98
4.10.5 Fits to Partial Regions of the Dalitz Plot
When examining the mass-squared projections for Fits 1-12 we see a small but con-
spicuous discrepancy between data and fit at around 1 GeV2 /c4 in both Kπ pro-
jections (there is a small deficit in the data compared to the fit). In the course of
understanding this feature (also known as “the dip”) we investigated a variety of
different matrix element models M. Several of these involved various K-matrix ap-
proaches (see Appendix A), none of which succeeded in modeling the data better
than the approach using Breit-Wigner resonances we have described so far in this
note. Fortunately, in the course of these studies we developed the ability to perform
partial fits over the Dalitz plot, motivated by the desire to understand exactly where
on the Dalitz plot the model discrepancy was the worst.
It is evident that most of the information relevant to extracting the relative am-
plitude rD and phase difference δD will be contained in the “+” shaped region that
is defined by the two K∗ bands on the Dalitz plot (i.e. the vertical and horizontal
bands of Figure 4.4(a)). In fact, the idea for determining δD described by Rosner
and Suprun [30] involves a simple counting experiment which basically compares the
yield of events in a region of overlap between K∗+ and K∗− bands to the yield in
the non-overlapping parts of the two bands. Thus, most of the information about δD
should be contained within the overlapping K∗ bands.
We should be able to study any systematic effect of our model’s inability to
describe areas of the Dalitz plot which are not in the K∗ signal region by separately
fitting the full Dalitz plot and the much smaller K∗ signal region with the same model
and looking for any changes to the extracted values of rD and δD. We also used other
choices for regions of the Dalitz plot to see if the fits for rD and δD change much.
The results of these studies are shown in Figures 4.34-4.36, 4.38-4.44, and Table 4.5.
See Figure 4.33 for a visual definition of each type of reduced fit region.
99
(a) (b)
(c) (d)
(e)
Black points (triangles) are includedCyan points (squares) are excluded
0.7 1.1 1.5 1.9
0.7
1.1
1.5
1.9
0.7 1.1 1.5 1.9
0.7
1.1
1.5
1.9
0.70.7 1.11.1 1.51.5 1.91.9
0.70.7
1.11.1
1.51.5
1.91.9
0.70.7 1.11.1 1.51.5 1.91.9
0.70.7
1.11.1
1.51.5
1.91.9
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
0.70.7 1.11.1 1.51.5 1.91.9
0.70.7
1.11.1
1.51.5
1.91.9
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
Figure 4.33: Fit Regions (a) K∗ bands (b) No φ 1.05 (c) No φ 1.15 (d) No φ 1.26 (e)“L” area.
100
To fit partial areas of the Dalitz plot, we use the same efficiency and background
functions which we determined from the whole Dalitz plot. However, only events
which are in the fitting region are used in the negative logarithmic likelihood function
L, and the efficiency and background are both set to zero outside the fitting region
(in order to get the correct normalization over the Dalitz plot).
Fits to the K∗ Bands
In our ‘K∗ bands’ fits we use K∗ bands which are defined as being bounded by
(mK∗ − ΓK∗)2 and (mK∗ + ΓK∗)2 along each of the m2Kπ axes (see Figure 4.33(a)).
These bands are twice as big in mKπ as the bands referred to in Reference [30]. The
K∗ bands region contains 334 D0 → K+K−π0 candidates.
Figure 4.34 (Fit 1x) shows the result of a partial fit to just the K∗ bands using
the nominal simple model that includes just K∗+, K∗−, plus a uniform non-resonant
component. We immediately notice that the strong phase difference and the K∗
amplitude ratio are about the same as in the nominal fit (Fit 1), the fit fraction of
the non-resonant component is greatly reduced (by almost a factor of two, although
this is not strange given the reduced fitting area), and there are no-longer any “dips”
around 1 GeV2 along the m2Kπ axes.
This seems to be a very good fit. Furthermore, the fact that δD and rD are
unchanged from their nominal values shows that our result should be rather insensitive
to the details of the modeling of the components which are outside the K∗ bands.
Figure 4.35 (Fit 2x) shows the effect of using the K∗s but ignoring the non-
resonant component. The quality of the fit is clearly worse than the nominal “K∗-
band” fit shown in Fit 1x — the summed negative log likelihood, L, increases by 60
and the fitted signal fraction is considerably lower than that for Fit 1x. This gives us
further confidence that a broad S-wave feature is present.
Figure 4.36 (Fit 3x) shows the effect of replacing the non-resonant contribution
101
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.34
:D
alit
zplo
tpro
ject
ions
for
Fit
1xsh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
102
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.35
:D
alit
zplo
tpro
ject
ions
for
Fit
2xsh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
103
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.36
:D
alit
zplo
tpro
ject
ions
for
Fit
3xsh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
104
0.4 0.6 0.8 1 1.2 1.4 1.6 1.80.4
0.6
0.8
1
1.2
1.4
1.6
1.8
±30o
0.4 0.6 0.8 1 1.2 1.4 1.6 1.80.4
0.6
0.8
1
1.2
1.4
1.6
1.8(a) (b)
0
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
mK-
π02 (GeV2/c4)
Figure 4.37: (a) The absolute value of the phase distribution of our nominal fit overthe entire Dalitz plot subtracted from the phase distribution of the nominal fit withthe φ simply removed (aφ is set to be 0). This shows the effect the φ resonance hason the overall phase at different points in phase space on the Dalitz plot. Removingonly a region where the φ has its peak is not enough to remove all regions of theDalitz plot whose phase is noticeably affected by the φ resonance. (b) The same plotas (a) except that areas outside the K∗ bands are excluded. Note that the phasedistribution is not affected by the φ in the vast majority of the K∗ bands region, soit is definitely appropriate to exclude the φ from any K∗ bands fit.
to the nominal fit with the κ. δD is much lower than in Fit 1x, but with a large
statistical uncertainty, while rD is higher. However, rD and δD are consistent with
Fit 3. This fit seems good for the most part, although the κs are less than 2 standard
deviations significant (FFκ+ = (7.6 ± 4.4)%, FFκ− = (10.1 ± 5.6)%).
Fits to Regions which Exclude the φ
An alternative set of fits we tried were fits that removed the φ resonance. The φ is
a narrow but significant part of the Dalitz Plot and does have an effect. We do not
believe these fits to be as useful as the K∗ bands fits, because simply removing the
φ from our fitting region is not enough to remove its effects on the Dalitz plot (see
Figure 4.37). However, by simply removing the φ we do not remove as many of our
105
statistics as in the K∗ bands fit.
Figure 4.38 (Fit 1n) shows the effect of excluding from the fit only the peak of the
φ resonance (Figure 4.33(b)). Specifically, we exclude m2K+K− between 1.02 GeV2 and
1.05 GeV2. The remaining region contains 613 D0 → K+K−π0 candidates. We again
use our simplest model consisting of just K∗+, K∗−, plus a uniform non-resonant
component, and we again obtain a good fit with the same value of rD and a fairly
unchanged, but slightly higher, value of δD. We believe that the φ still affects the
region of this plot, so leaving it out of the fit is likely causing the shift in δD.
Figure 4.39 (Fit 1np) uses the same fit region as Fit 1n. We add the φ back in to
go along with K∗+, K∗−, plus a uniform non-resonant component. We obtain a good
fit and δD is more in line with Fit 1 than Fit 1n was, although rD decreases slightly.
The fit fraction for the φ is not very significant, as expected.
Figure 4.40 (Fit 1na) shows the effect of excluding m2K+K− less than 1.15 GeV2
from the fit (Figure 4.33(c)). We fit with the same resonances as in Fit 1n. This
region contains 573 D0 → K+K−π0 candidates. δD increases compared to Fit 1,
perhaps for the same reason as in Fit 1n, and rD decreases slightly when compared
to Fit 1.
Figure 4.41 (Fit 1nap), shows a fit which is the same as Fit 1na but with the φ
added. This fit did not converge in the fitter at first with the default starting values.
When I changed the starting values for amplitudes and phases of the fit to those in
Fit 1, the fitter did converge. The fit fraction for the φ is not significant, as expected.
δD is now a few degrees less than in Fit 1na, and rD decreases further.
Figure 4.42 (Fit 1nb) shows the effect of excluding m2K+K− less than 1.26 GeV2
from the fit (Figure 4.33(d)). We fit with the same resonances as in Fit 1n. This
region contains 552 D0 → K+K−π0 candidates. rD is slightly lower than Fit 1, and
δD is still greater than the δD in Fit 1, but still consistent with Fit 1.
Figure 4.43 (Fit 1nbp) uses the same fit region as Fit 1nb. We add the φ back in
106
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.38
:D
alit
zplo
tpro
ject
ions
for
Fit
1nsh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
107
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.39
:D
alit
zplo
tpro
ject
ions
for
Fit
1np
show
ing
bot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
108
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.40
:D
alit
zplo
tpro
ject
ions
for
Fit
1na
show
ing
bot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
109
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.41
:D
alit
zplo
tpro
ject
ions
for
Fit
1nap
show
ing
bot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
110
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.42
:D
alit
zplo
tpro
ject
ions
for
Fit
1nb
show
ing
bot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
111
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.43
:D
alit
zplo
tpro
ject
ions
for
Fit
1nbp
show
ing
bot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
112
to go along with K∗+, K∗−, plus a uniform non-resonant component. rD is the same
as in Fit 1nb, but δD is more consistent with what is found in Fit 1. However, the φ
fit fraction is not significant, as expected.
Fits to the “L” Region
Figure 4.44 (Fit 1s) excludes events for which both m2K+π0 and m2
K−π0 are greater
than 1.10 GeV2 (Figure 4.33(e)). This is an “L” shaped region in m2K+π0 and m2
K−π0
which contains 549 D0 → K+K−π0 candidates. This region is a compromise between
the K∗ bands fit and the no φ fit regions. We again use just K∗+, K∗−, and a uniform
non-resonant component. We again obtain a good fit. It is interesting that these fit
constraints remove the φ also, but δD and rD are fairly unchanged compared to Fit
1.
4.10.6 Floating mκ, Γκ Fit to the Entire Dalitz Plot
Finally, we do a fit which allows the mass and width of the κ to float, since the mass
and width of the κ are not well determined by other experiments.
Figure 4.45 (Fit 3f) shows a fit to the entire Dalitz plot with the resonance set:
{K∗+, K∗−, κ+, κ−, and φ} where the κ mass and width are allowed to float. The fit
to the data is good. The fit prefers mκ = 855 ± 15 MeV and Γκ = 251 ± 48 MeV.
The relative phase δD between K∗− and K∗+ falls to 306◦, and rD remains consistent
with Fit 1.
113
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.44
:D
alit
zplo
tpro
ject
ions
for
Fit
1ssh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
114
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
4.45
:D
alit
zplo
tpro
ject
ions
for
Fit
3fsh
owin
gbot
hth
efit
(red
curv
e)an
dth
ebin
ned
dat
asa
mple
.
115
4.11 Systematic Errors for rD and δD in our
Nominal Fit
We consider systematic errors from experimental sources and from the decay model
separately. Our general procedure is to change some aspect of the analysis and
interpret the change in the values of the amplitude ratio rD and phase difference
δD as an estimate of the associated systematic uncertainty. Contributions to the
experimental systematic uncertainties arise from our models of the background, the
efficiency, the signal fraction, and the event selection. The model systematic error
arises from uncertainty in the choice of resonances used to fit the Dalitz plot.
Part of our systematic error is the variation of δD and rD based on different models
used for our matrix element. We also expect that changing the fitting region will be
part of our systematic error. Also, fits that have resonance choices which lead to very
unreasonable fit fractions and amplitudes should also be excluded. This leaves the
set of fits {1, 1n, 1np, 1na, 1nap, 1nb, 1nbp, 1x, 1s, 3, 3f, 4, 5, 11, 12}. Within these
fits, we believe that fits that are in “no φ” regions which did not include the φ may
still have lingering interference effects from the φ which affect the K∗s. So we will
define fits {1, 1x, 1np, and 1s} within the category of changing the fitting region on
the Dalitz plot, and we characterize fits {1, 3, 3f, 4, 5, 11, 12, 1np, 1nap, 1nbp} to
be within the category of changing our model.
We can also do a fit to (m2K−π0 , m2
K+π0) rather than (m2K+π0 , m2
K+K−). In the
(m2K−π0 , m2
K+π0) fit, the third Dalitz plot variable, m2K+K− , is calculated from m2
K−π0
and m2K+π0 . Because we do not enforce an energy-momentum constraint on all three
invariant-mass squared variables measured, the distribution for calculated m2K+K− is
not the same as that for measured m2K+K− . Unfortunately, this difference we found is
large enough such that we believe the φ resolution is much worse in calculated m2K+K−
[Γ ≈ 30 MeV] than measured m2K+K− [Γ ≈ 5 MeV]. However, we repeat the Dalitz fit
116
for the new choice of Dalitz plot variables as a means to understand the systematics
of our choice of invariant mass-squared variables. This fit used a background and
efficiency parametrization based on a best fit to m2K−π0 and m2
K+π0 . We fixed the
signal fraction to 81.8% in the (m2K−π0 , m2
K+π0) fit and the (m2K+π0 , m2
K+K−) fit for
the systematic comparison. We did not expect the φ to be fit very well, and its
amplitude and phase are inconsistent between the two fits.
Table 4.8 lists our systematic checks and systematic uncertainties. Systematic
Check 1 represents our modeling error, and the other systematic errors represent our
experimental error. We take the square root of the sample variance of the amplitudes
and phases from the nominal result (Fit 1) compared to the results in the series of
variation fits used for systematics checks as a measure of the systematic uncertainty
[51]. We find systematic errors of ±2.9◦(experimental) ± 10.6◦(model) for δD and
±0.016(experimental) ± 0.038(model) for rD. Adding systematic errors in this way
results in a model systematic error for δD that is less than the difference in δD when
comparing Fit 1 to Fit 3. We add the experimental and model systematic uncertainty
in quadrature to obtain the total systematic uncertainty shown in Table 4.8.
Our systematic error is dominated by the model dependence, and the largest
deviations from the nominal fit were observed in the series of fits where we replaced the
non-resonant contribution with the κ±. If fits including a κ± resonance are removed
from consideration, then the systematic errors on δD and rD decrease from ±11◦
and ±0.04 to ±8◦ and ±0.03, respectively, and the remaining systematic uncertainty
is dominated by fits including the K0(1430)±. The largest experimental systematic
uncertainties are ±8◦ for δD when allowing the background parameters to float, and
±0.05 for rD when allowing the efficiency parameters to float, as described below.
Table 4.9 provides details about our systematic checks. “Lo Acceptance” refers to
tighter selection criteria (less events pass), whereas “Hi Acceptance” refers to looser
selection criteria (more events pass). Generally the signal fraction is allowed to float
117
#Syst
emat
icC
hec
kU
nce
rtai
nty
inδ D
Unce
rtai
nty
inr D
1A
ddin
got
her
reas
onab
lere
sona
nces
±11◦
±0.0
42
Cha
ngin
gth
efit
ting
regi
onon
theDP
±2◦
±0.0
23
Fit
ting
(m2 K
−π
0,m
2 K+
π0)
rath
erth
an(m
2 K+
π0,m
2 K+
K−)
±1◦
(<)
±0.0
34a
/4b
Var
iati
onin
π0
crit
eria
(γγ
mas
sac
cept
ance
,E
nerg
y)±4
◦/±3
◦±0
.02/
±0.
01(<
)5
Var
iati
onin
D∗
frac
tion
alm
omen
tum
requ
irem
ent
±5◦
±0.0
16
RIC
Hse
lect
ion
crit
eria
*±5
◦±0
.03
7m
K+
K−
π0
acce
ptan
ce±2
◦±0
.01
8Δ
Mac
cept
ance
±2◦
±0.0
1(<
)9
Var
ysi
zeof
the
ΔM
vs.
mK
+K
−π
0w
indo
w±2
◦±0
.02
10V
ary
dE/d
xPar
amet
ers*
*±1
◦±0
.01
11V
ary
the
Ori
gina
lSi
gnal
Frac
tion
±5◦
±0.0
112
Var
yth
eO
rigi
nalSi
gnal
Frac
tion
via
Pen
alty
Ter
m±1
◦±0
.01
(<)
13V
ary
the
back
grou
ndpa
ram
eter
s±8
◦±0
.01
14V
ary
the
effici
ency
para
met
ers
±4◦
±0.0
515
Var
yth
ew
idth
ofth
eφ
byal
low
ing
itto
float
±1◦
(<)
±0.0
1Tot
alSy
stem
atic
Err
or(A
dded
asin
Ref
eren
ce[5
1]):
±11◦
±0.0
4
Tab
le4.
8:Sum
mar
yof
syst
emat
icch
ecks
and
syst
emat
icer
rors
.“<
”m
eans
asm
all
chan
ge.
*R
ICH
sele
ctio
ncr
iter
iaar
euse
dw
hen
RIC
Hin
fois
avai
lable
for
kaon
sw
ith
mom
entu
mp K
>50
0M
eV.**
dE
/dx
sele
ctio
ncr
iter
iaar
euse
dw
hen
p K<
500
MeV
orw
hen
RIC
Hin
fois
not
avai
lable
for
kaon
sw
ith
mom
entu
mp K
>50
0M
eV.
118
in the fitter as it was for the nominal fit, unless otherwise stated.
Systematic checks 1-3 are described above.
In systematic checks 4a and 4b, we could only tighten our selection criteria due
to pre-existing limits placed to make the data accessible.
Systematic checks 5, 7, 8, 9, and 10 are simple variation checks where we increase
and decrease our selection criteria and see how it affects our results.
Systematic check 6 is a cut on a subtraction of how likely the particle is to be a
kaon vs. how likely it is to be a pion. Since 0 is the point where a kaon hypothesis is as
likely as a pion hypothesis, it was hard to determine an amount of change for a test of
systematics. Instead, our guideline was to choose variations which did not change the
number of events accepted drastically from the 709 accepted in the nominal fit. 680
events were accepted for the “Lo Acceptance” RICH systematic fit and 783 for the
“Hi Acceptance” RICH systematic fit. The number of events accepted decreased by
4.1% for tighter selection criteria and increased by 10.4% for looser selection criteria.
Systematic check 11 is a check to see how much a change in the signal fraction
affects the fitter. The signal fraction determined from the ΔM and mK+K−π0 fits is
81.8% ± 6.3%(stat.) ± 2.8%(syst.), while the signal fraction from the nominal Dalitz
fit (Fit 1) is 86 ± 6%. The central values differ by about 0.7 standard deviations.
In this check, we simply varied the signal fraction from the original signal fraction to
different values which could be reasonable given the statistical error.
Systematic check 12 adds a term to the negative logarithmic likelihood function
L which allows the signal fraction to float but penalizes L depending on how far the
signal fraction floats from the original signal fraction f0 = 81.8% ± 6.3%(stat.). This
term looks like:
Penalty Term = (f − f0
σf
)2. (4.10)
119
#Syst
emat
icC
hec
kO
rigi
nal
Val
ue
Lo/
HiA
ccep
tance
Var
iation
4aV
aria
tion
inγγ
(π0)
mas
sac
cept
ance
|σ π0|<
2.5
std.
dev.
|σ π0|<
2st
d.de
v.,
N/A
20%
4bV
aria
tion
inπ
0E
nerg
ycr
iter
ia10
0M
eV>
110
MeV
,N
/A10
%
5V
aria
tion
inD
∗fr
acti
onal
mom
entu
mre
quir
emen
t0.
50.
55,0.
4510
%
6R
ICH
crit
eria
*0
-10,
10N
/A
7m
K+
K−
π0
acce
ptan
ce1.
84to
1.89
GeV
1.84
5to
1.88
5G
eV,
1.83
5to
1.89
5G
eV20
%
8Δ
Mac
cept
ance
|ΔM|<
.001
0G
eV|Δ
M|<
.000
8,|Δ
M|<
.001
220
%
9V
ary
size
ofth
eΔ
Mvs
.m
K+
K−
π0
win
dow
1.84
to1.
89G
eV|Δ
M|<
.001
0G
eV
1.84
25to
1.88
75G
eV|Δ
M|<
.000
9,1.
8375
to1.
8925
GeV
|ΔM|<
.001
1
10%
ofea
ch.
Cha
nge
inar
eaof
win
dow
:A
bout
20%
10V
ary
dE/d
xPar
amet
ers*
*|σ K
|<2.
5st
d.de
v.|σ K
|<2
std.
dev.
,|σ K
|<3
std.
dev.
20%
#Syst
emat
icC
hec
kO
rigi
nal
Val
ue
New
Val
ue(
s)V
aria
tion
11V
ary
the
Sign
alFr
acti
on81
.8%
±6.
3%75
.5%
,88
.1%
1st
d.de
v.ab
ove
and
belo
wor
igin
alsi
gnal
frac
tion
12V
ary
the
Sign
alFr
acti
onvi
aPen
alty
Ter
m81
.8%
±6.
3%84
.0%
±4.
3%0.
4st
d.de
v.
13V
ary
the
back
grou
ndpa
ram
eter
svi
aB
ackg
roun
dC
ovar
ianc
eM
atri
xSe
eTab
le4.
4Se
eTab
le4.
10Se
eTab
le4.
10
14V
ary
the
effici
ency
para
met
ers
See
Tab
le4.
3Se
eTab
le4.
11Se
eTab
le4.
11
15V
ary
the
wid
thof
the
φby
allo
win
git
toflo
atΓ
φ=
4.26
MeV
Γφ
=5.
30M
eV24
.4%
Tab
le4.
9:D
etai
lsab
out
syst
emat
icch
ecks.
*R
ICH
sele
ctio
ncr
iter
iaar
euse
dw
hen
RIC
Hin
fois
avai
lable
for
kaon
sw
ith
mom
entu
mp K
>50
0M
eV.**
dE
/dx
sele
ctio
ncr
iter
iaar
euse
dw
hen
p K<
500
MeV
orw
hen
RIC
Hin
fois
not
avai
lable
for
kaon
sw
ith
mom
entu
mp K
>50
0M
eV.
120
For the purpose of systematic checks 11 and 12, we ignore the comparatively small
(± 2.8%) systematic error on the signal fraction.
Systematic check 13 adds a term to the function L which allows the background
parameters to float up to one standard deviation away from their original values, but
penalizes L depending on how far the background parameters float from the original
parameters in Table 4.4. This term looks like:
Background Covariance Term = Σij(Bi − Bio)Vij(Bj − Bjo), (4.11)
where Bi is one of the background parameters and Bio is the original value of that
parameter, and where V is the background covariance matrix which relates variations
in the parameter Bj to variations in Bi. The fit requires that the resonant background
parameters can only vary by 1 standard deviation.
Table 4.10 shows the best-fit parameters for the background shape when the back-
ground parameters are allowed to float, with a background covariance matrix penalty
term.
B0 By Bz Byy Byz
1 (fixed) 20.58 ± 0.1355 11.96 ± 0.2474 92.66 ± 0.2343 -88.91 ± 0.2687Bzz Byyy Byyz Byzz Bzzz
10.22 ± 0.4594 -43.16 ± 0.4716 18.95 ± 0.4086 9.040 ± 0.5994 0.4483 ± 1.260BK∗+ BK∗− Bφ
0.04942 ± 5.208x10−5 .006606 ± .001932 0.03932 ± 3.311x10−7
Table 4.10: Best-fit parameters for the background shape when the background pa-rameters are allowed to float.
Systematic check 14 is to change the fit by allowing the efficiency parameters to
all float within 1 standard deviation of their original values. We did not use the
efficiency covariance matrix to further constrain this fit.
Table 4.11 shows the best-fit parameters for the efficiency shape when the effi-
121
ciency parameters are allowed to float.
E0 Ey Ez Eyy Eyz
1 (fixed) -4459 ± 149 1458 ± 92 3261 ± 57 2562 ± 188
Ezz Eyyy Eyyz Eyzz Ezzz
-46.54 ± 36.23 -597.9 ± 31.9 -830.5 ± 70.0 -640.2 ± 261.8 -81.78 ± 13.12
Table 4.11: Best-fit parameters for the efficiency shape when the efficiency parametersare allowed to float.
4.12 Branching Ratio Cross-check
As a cross-check of our data, we estimate the branching ratio of D0 → K+K−π0 from
our data and compare it to the published value. Branching ratio measurements are
not the focus of this analysis, so systematic errors have not been investigated. Based
on our mK+K−π0 fit, we have a total of 627 ± 30 signal events. We measured our
average efficiency3 for finding the decay chain D∗+ → π+slowD0; D0 → K+K−π0 from
the signal Monte Carlo (discussed in Section 4.7) to be ε = (5.83 ± 0.09)%. From this,
we calculated how many D0 → K+K−π0 we expected in our data sample (our yield
divided by ε). We may estimate the total number of D0s expected from continuum
D∗+s in our data sample, based on the integral cross-section for continuum D∗+
production near√
s = 10.6 GeV [52]: σ(e+e− → D∗+X) = (583 ± 8 ± 33 ± 14) pb,
where the fourth error stems from external branching fraction uncertainties. By
dividing our yield by our efficiency to determine the number of D0 → K+K−π0
decays in our sample, taking that result and dividing by the total number of D0s we
expect from continuum D∗+s, and knowing BR(D∗+ → π+s D0) = (67.7 ± 0.5)% [1],
we estimate BR(D0 → K+K−π0) = (0.30±0.02)%, which is significantly higher than
3The average efficiency ε is simply the number of MC events we are able to reconstruct (4077.3± 64.43) as D∗+ → π+
slowD0; D0 → K+K−π0, using the same computer code we use to reconstructdata events, divided by the total number of generated signal (D∗+ → π+
slowD0; D0 → K+K−π0)MC events (69922). This is the same sample used to create the MC Dalitz Plot in Section 4.7.
122
the previous measurement of BR(D0 → K+K−π0) = (0.13 ± 0.04%) [1, 31].
Combining our fit fractions with known values for BR(K∗± → K±π0) and BR(φ →K+K−) [1], we also estimate branching ratios of the resonant decay modes. We find
BR(D0 → K∗+K−) = (0.38 ± 0.04)%, BR(D0 → K∗−K+) = (0.10 ± 0.02)%, and
BR(D0 → φπ0) = (0.084 ± 0.012)%. These branching ratios are consistent with
published measurements [1, 32–34] and are listed in Table 4.12.
A recent4 measurement of BR(D0 → K+K−π0) = (0.334±0.004±0.006±0.012)%
by the BaBar Collaboration [53] is consistent with our estimate from CLEO III data.
Decay Mode Published BR Estimated BRD0 → K∗+K− (0.37 ± 0.08)% (0.38 ± 0.04)%D0 → K∗−K+ (0.20 ± 0.11)% (0.10 ± 0.02)%
D0 → φπ0 (0.076 ± 0.005)% (0.084 ± 0.012)%D0 → K+K−π0 non-resonant No Measurement (≈ 0.09)%
Table 4.12: Comparison of estimated branching ratios from this analysis to publishedbranching ratios [32–34, 48] based on our nominal Dalitz plot fit. We accounted forthe fact that K∗± decays to K±π0 only one-third of the time, and φ decays to K+K−
only about one-half of the time.
4.13 U-spin Symmetry Check
The U -spin symmetry interchanging s and d quarks predicts the following for D0
decays to a pseudoscalar meson and a vector meson [54]:
A(D0 → π+ρ−) = −A(D0 → K+K∗−) (4.12)
and
A(D0 → π−ρ+) = −A(D0 → K−K∗+), (4.13)
4This measurement was published very recently (on 15 November 2006) so, due to time con-straints, this chapter has not been updated to include this result as “current knowledge”. However,the journal reference is cited in the References.
123
where A is the respective dimensionless invariant amplitude for each decay. Dividing
Equation 4.12 by Equation 4.13, and assuming that A(K∗+ → K+π0) = A(K∗− →K−π0), gives:
A(D0 → π+ρ−)
A(D0 → π−ρ+)=
aK∗−K+
aK∗+K−ei(δK∗−K+−δK∗+K− ). (4.14)
Assuming a phase convention such that a phase difference of 0◦ indicates maximal
destructive interference between ρ− and ρ+, and assuming A(ρ+ → π+π0) = A(ρ− →π−π0), we can use the recently published results of a Dalitz plot analysis of D0 →π+π−π0 [40] to evaluate the left hand side of Equation 4.14:
(0.65 ± 0.03 ± 0.02)ei(356◦±3◦±2◦)
which may be compared to the right hand side of Equation 4.14 which comes from
this analysis:
(0.52 ± 0.05 ± 0.04)ei(332◦±8◦±11◦).
We do not expect these results to necessarily be consistent, since U -spin symmetry
does not always make correct predictions [54]. However, any symmetry-breaking
effects are important to identify.
4.14 The Effect of our Uncertainties on γ
It is interesting to look at how the uncertainties in our measurements of rD and δD
affect the measurement of γ. For convenience we repeat the equations (discussed in
124
Section 1.6) which may be used to determine γ from δD and rD:
A[B− → K−(K∗+K−)D] = |ABAD|(1 + rBrDei(δB+δD−γ)) (4.15)
A[B− → K−(K∗−K+)D] = |ABAD|eiδD(rD + rBei(δB−δD−γ)) (4.16)
A[B+ → K+(K∗−K+)D] = |ABAD|(1 + rBrDei(δB+δD+γ)) (4.17)
A[B+ → K+(K∗+K−)D] = |ABAD|eiδD(rD + rBei(δB−δD+γ)), (4.18)
If we calculated γ via Equations 4.15 or 4.17, the size of the uncertainty in γ due
to uncertainty in δD is simply equal to the size of the uncertainty in δD, and does not
depend on rD. However if we use Equations 4.16 or 4.18, the uncertainty in γ due to
the uncertainty in rD or δD is more difficult to know. We may define:
A′Eq. 4.16 =
A[B− → K−(K∗−K+)D]
|ABAD| (4.19)
A′Eq. 4.18 =
A[B+ → K+(K∗+K−)D]
|ABAD| . (4.20)
Then we can determine the real parts of the partial derivatives of γ to get an idea
of how the uncertainty in γ is affected by the uncertainty in δD or rD when using
Equations 4.16 or 4.18:
Re
(∂γ
∂δD
)Eq. 4.16
= Re
[A′
Eq. 4.16
A′Eq. 4.16 − rDeiδD
]− 1 (4.21)
Re
(∂γ
∂rD
)Eq. 4.16
= Im
[1
A′Eq. 4.16 e−iδD − rD
](4.22)
Re
(∂γ
∂δD
)Eq. 4.18
= Re
[A′
Eq. 4.18
rDeiδD − A′Eq. 4.18
]+ 1 (4.23)
Re
(∂γ
∂rD
)Eq. 4.18
= Im
[1
rD − A′Eq. 4.18 e−iδD
], (4.24)
where Im[x] is the imaginary part of x and Re[x] is the real part of x. The uncertainty
125
in γ in these cases also depends on the values of δD and rD themselves. Imaginary
parts of the partial derivatives do not affect γ, these represent effects which rescale
the amplitudes but do not affect the phase.
The value of γ from direct measurements (not including the recent measurement
by Belle in Reference [12]) is γ = (63+15−12)
◦ [1]. We expect that the effect on the
uncertainty of γ due to the uncertainties of δD and rD from this analysis is comparable
to the current errors from direct measurements of γ.
4.15 Conclusion
In conclusion, we have examined the resonant substructure of the decay D0 →K+K−π0 using the Dalitz plot analysis technique. We observe resonant K∗+K−,
K∗−K+, and φπ0 contributions. We also observe a significant S-wave modeled as a
κ±K∓ or a non-resonant contribution. Other models of the S-wave (including the K-
matrix formalism discussed in Appendix A) were not used but should be considered
and tested in future, higher-statistics analyses. We determine δD = 332◦ ± 8◦ ± 11◦
and rD = 0.52 ± 0.05 ± 0.04. The measurements of δD and rD seem robust against
variations in signal selection as well as background and efficiency determination.
126
5 D0 → K+K−π0 in CLEO-c
The final phase of the CLEO experiment began June 2003 and is called CLEO-c.
CLEO-c uses mostly the same detector as CLEO III, but replaces the silicon vertex
detector with a small drift chamber [25]. CLEO-c is expected to record about 1500
resonant decays of the type useful for determining the strong phase δD, an amount
that will allow a determination of cos(δD) with an uncertainty of ±0.27 or better [30].
A preliminary look at D0 → K+K−π0 in CLEO-c is presented here.
5.1 Effects of Quantum Correlations
To tag the flavor of the D0(D0), we have to take advantage of the fact that at
√s ≈ 3.77 GeV a D0 can only be produced by the process e+e− → D0D
0. If we
can cleanly tag the other side of the decay to a mode which primarily decays from a
D0, we will be able to say with confidence that the side we are interested in (which
decays to K+K−π0) is a D0, and vice versa. However, the results of our fit may
be modified by the effects of quantum correlations of the C = −1 initially produced
state1 e+e− → D0D0. According to Reference [55], the rates, R, for a decay which
has one side decaying to a flavor-tag mode f and the other side decaying to a non-CP
eigenstate f ′ or its conjugate mode f′are:
Rf ′,f = BR(D0 → f ′)BR(D0 → f)(1 + r2f ′r2
f − rf ′rfv−f ′,f ) (5.1)
Rf′,f
=BR(D0 → f
′)BR(D0 → f)(r2
f ′ + r2f − rf ′rfv
+f ′,f )
r2f ′
, (5.2)
1C = Charge Conjugation eigenvalue
127
where f is the conjugate of f . The rate for a decay which has one side decaying to a
flavor-tag mode and the other side decaying to a CP = +1 eigenstate S+ is:
RS+,f = BR(D0 → S+)BR(D0 → f)(1 + r2f + rfzf ). (5.3)
In both cases:
rj =
∣∣∣∣〈j|D0〉〈j|D0〉
∣∣∣∣ (5.4)
δj = − arg
(〈j|D0〉〈j|D0〉
)= −(δstrong + δweak + 180◦) (5.5)
v±f ′,f = 2 cos δf ′ cos δf ± 2 sin δf ′ sin δf (5.6)
zf = 2 cos δf , (5.7)
and CP is assumed to be conserved for D decays. The conjugate states j and j are
defined such that rj ≤ 1.
For the resonant modes on our Dalitz plot, we would like to know how their
rates are affected by quantum correlations of the original D0D0
state. We have
two conjugate non-CP eigenstate modes D0 → K∗+K− and D0 → K∗−K+, and a
CP = +1 mode D0 → φπ0. In order to calculate rK∗K and δK∗K we simply use
the information from our CLEO III analysis where the D0s were uncorrelated (see
Chapter 4)2. We use:
rK∗K =
∣∣∣∣〈K∗−K+|D0〉〈K∗+K−|D0〉
∣∣∣∣ = 0.52 ± 0.05 ± 0.04, (5.8)
2Our phase convention for δD is consistent with the phase convention of δstrong in Reference [55].
128
and
δK∗K = − arg
(〈K∗−K+|D0〉〈K∗+K−|D0〉
)= −(δstrong + δweak + 180◦)K∗K
= −(332◦ + 0◦ + 180◦) ± 8◦ ± 11◦
= −512◦ ± 8◦ ± 11◦
= 208◦ ± 8◦ ± 11◦. (5.9)
For this calculation we will assume our tag mode is f = K−π+. rKπ is known for this
mode to be [55]:
rKπ =
∣∣∣∣〈K+π−|D0〉〈K−π+|D0〉
∣∣∣∣ = 0.0612 ± 0.0015, (5.10)
and we assume
δKπ = − arg
(〈K+π−|D0〉〈K−π+|D0〉
)= −(δstrong + δweak + 180◦)Kπ
= −(0◦ + 180◦ + 180◦)
= −360◦
= 0◦. (5.11)
We may then calculate the ratios of expected rates for correlated D0s in terms of
the uncorrelated rates (the branching ratios, BR) from Equations 5.1 through 5.11
and find:
RK∗+K−,Kπ
RK∗−K+,Kπ
= (0.87 ± 0.03) · BR(D0 → K∗+K−)
BR(D0 → K∗−K+), (5.12)
RK∗+K−,Kπ
Rφπ0,Kπ
= (0.94 ± 0.06) · BR(D0 → K∗+K−)
BR(D0 → φπ0), (5.13)
129
and
RK∗−K+,Kπ
Rφπ0,Kπ
= (1.08 ± 0.07) · BR(D0 → K∗−K+)
BR(D0 → φπ0). (5.14)
where the errors are propagated from the total errors on rK∗K and δK∗K . We do not
include the error on rKπ since the error is relatively small, and we assume δKπ to be
identically zero [55].
These results may vary if another tag mode f is used, although this calculation
is representative of how the other tag modes (K−π+π0 and K−π+π+π−) will affect
the results.
5.2 Data Sample
This analysis uses an integrated luminosity of 281 pb−1 of e+e− collisions at√
s ≈3.77 GeV provided by the Cornell Electron Storage Ring (CESR). The data were
collected with the CLEO-c detector [25]. Charge conjugation is implied as in Chapter
4. Flavor tag modes used in this analysis include D0 → K−π+, D0 → K−π+π0, and
D0 → K−π+π+π−.
The analysis uses a standard set of selection criteria for the charged kaons and
neutral pions. The detected charged particle tracks must reconstruct to within 5 cm of
the expected interaction point along the beam pipe and within 5 mm perpendicular
to the beam pipe. Kaon candidates are required to have momenta 50 ≤ pK ≤2000 MeV/c. The cosine of the angle θ between a track and the nominal beam axis
must be between −0.93 and 0.93 in order to assure that the particle is in the fiducial
volume of the detector.
Candidate kaon tracks that have momenta greater than or equal to 700 MeV/c are
selected based on combined information from the Ring Imaging Cherenkov (RICH)
130
detector and dE/dx if at least three photons associated with the track are detected.
The combined likelihood used in this case is that defined by Equation 3.6 and the
selection requirement is Δχ2 ≤ 0. Candidate kaon tracks without RICH information
or with momentum below 700 MeV/c are required to have specific energy loss in the
drift chamber within 3 standard deviations of that expected for a true kaon (|σK | < 3).
We still require Δχ2 ≤ 0, except that in the case where only dE/dx information is
used, we ignore the RICH terms in Equation 3.6. The RICH is also not used for
| cos θ| > 0.8, to avoid an inefficiency near the edge of the RICH for kaons.
The π0 candidates are reconstructed from all pairs of electromagnetic showers that
are not associated with charged tracks. The two-photon invariant mass is required to
be within 3 standard deviations of the known π0 mass. To improve the resolution on
the π0 three-momentum, the γγ invariant mass is constrained to the known π0 mass.
5.3 Event Candidates
The D0 candidate beam-constrained mass mbc ≡ 1c2
√E2
beam − p2c2 and ΔE ≡ E −Ebeam are calculated for each candidate, where Ebeam =
√s
2is the electron (or positron)
beam energy, and p and E are the reconstructed momentum and energy, respectively,
of the decay particles. The distributions of mbc and ΔE are shown in Figure 5.1.
We select a signal region defined by 1.86 < mbc < 1.87 GeV/c2 and |ΔE| <
30 MeV/c2, which contains 211 D0 → K+K−π0 candidates. There is very little
background thanks to the clean D-meson decays at√
s ≈ 3.77 GeV.
5.4 CLEO-c Dalitz Plot Fit
Figure 5.2 shows the Dalitz plot distribution (a scatter plot of m2K+π0 vs. m2
K−π0)
for the D0 → K+K−π0 candidates in our signal region satisfying the requirements
131
h4Entries 246
Mean 1.865
RMS 0.003834
1.85 1.86 1.87 1.880
4
8
12
16
20
24Entries 246
Mean 1.865
RMS 0.003834
0π
-K+Kbcm (GeV/c2)
Entries 246
Mean -0.003417
RMS 0.02654
0
5
10
15
20
25 Entries 246
Mean -0.003417
RMS 0.02654
0π
-K+KE (MeV)Δ
Even
ts p
er 4
00 k
eV/c
2
Even
ts p
er 2
MeV
40 80-40-80 0
Signal Region
Figure 5.1: Plots of mbc and ΔE for all events passing the selection criteria in Section5.2, with the signal region within the dashed (brown) lines of both figures.
described in Sections 5.2 and 5.3.
Fitting the Dalitz plot is done by the procedure outlined in Chapter 4, except
that we assume that our efficiency and background functions are uniform (E(y, z) =
1, B(y, z) = 1) for our preliminary look at this data. Figure 5.3 shows the results
of the simplest possible reasonable fit to the data, Fit 1-c. In this fit the matrix
element M is composed of the three nominal resonances (K∗+, K∗−, and φ) plus a
uniform non-resonant (NR) component. The amplitudes, phases, and fit fractions
are shown in Table 5.1. The signal fraction was allowed to float in the fit, and is
f1−c = (92 ± 10)%. rD and δD are consistent with our nominal fit (Fit 1) result
from Chapter 4, but keep in mind that the amplitudes and phases of Fit 1-c are not
adjusted for effects of quantum correlations.
Figure 5.4 shows the results of Fit 3-c, which has a matrix element M composed
of K∗+, K∗−, κ+, κ−, and φ resonances. The κ has the mass and width shown in
Table 4.2. The amplitudes and phases are shown in Table 5.2. The signal fraction
132
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
mK-
π02 (GeV2/c4)
mK
+π
02 (GeV
2 /c4 )
Figure 5.2: Dalitz plot of K+K−π0 candidates.
Resonance Amplitude Phase Fit FractionK∗+ 1.00 (fixed) 0◦ (fixed) (50.2 ± 5.6)%K∗− 0.52 ± 0.07 333◦ ± 14◦ (13.8 ± 3.6)%φ 0.63 ± 0.07 311◦ ± 18◦ (16.0 ± 3.4)%
Non-Resonant 4.87 ± 0.93 245◦ ± 9◦ (29.2 ± 7.5)%Total Fit Fraction: (109.2 ± 10.6)%
Table 5.1: Amplitude, phase, and fit fraction results for Fit 1-c with statistical errors.
133
11.
52
2.5
305101520253035
0.4
0.6
0.8
11.
21.
41.
61.
82
051015202530
0.4
0.6
0.8
11.
21.
41.
61.
82
05
1015202530354045 Events per 0.0583 GeV2/c
4
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.07 GeV2/c
4
Events per 0.0583 GeV2/c
4
Sig
nal
Bac
kgro
un
dTo
tal
Fig
ure
5.3:
Dal
itz
plo
tpro
ject
ions
for
Fit
1-c.
134
Events per 0.0583 GeV2/c
4
mK
+K-2 (G
eV2 /c
4 ) m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.07 GeV2/c
4
Events per 0.0583 GeV2/c
4
11.
52
2.5
305101520253035
0.4
0.6
0.8
11.
21.
41.
61.
82
051015202530
0.4
0.6
0.8
11.
21.
41.
61.
82
05
1015202530354045Si
gn
alB
ackg
rou
nd
Tota
l
Fig
ure
5.4:
Dal
itz
plo
tpro
ject
ions
for
Fit
3-c.
135
was allowed to float in the fit, and is f3−c = (91±7)%. rD and δD are consistent with
our κ fit (Fit 3) result from Chapter 4, although rD is higher in Fit 3-c. Keep in mind
that the amplitudes and phases of Fit 3-c are not adjusted for effects of quantum
correlations.
Resonance Amplitude Phase Fit FractionK∗+ 1.00 (fixed) 0◦ (fixed) (39.3 ± 4.1)%K∗− 0.62 ± 0.10 311◦ ± 16◦ (15.4 ± 3.9)%φ 0.68 ± 0.08 332◦ ± 21◦ (17.1 ± 3.7)%κ+ 2.91 ± 0.58 194◦ ± 14◦ (28.7 ± 6.9)%κ− 1.34 ± 0.49 217◦ ± 25◦ ( 6.2 ± 2.8)%
Total Fit Fraction: (106.7 ± 10.1)%
Table 5.2: Amplitude, phase, and fit fraction results for Fit 3-c with statistical errors.
For the purpose of checking if quantum correlations have any effect, we calculate
ratios of fit fractions from CLEO-c and CLEO III data and compare them to what
is predicted from Equations 5.12 through 5.14, as shown in Table 5.3.
Ratio (X) Predicted Xcorrelated
Xuncorrelated
Xcorrelated(Fit 1-c)Xuncorrelated(Fit 1)
Xcorrelated(Fit 3-c)Xuncorrelated(Fit 3)
FF (K∗+)FF (K∗−)
0.87 ± 0.03 0.97 ± 0.33 0.68 ± 0.24FF (K∗+)
FF (φ)0.94 ± 0.06 1.01 ± 0.28 0.77 ± 0.22
FF (K∗−)FF (φ)
1.08 ± 0.07 1.04 ± 0.41 1.12 ± 0.46
Table 5.3: Comparison of predicted to measured ratios of fit fractions when comparingCLEO-c data (correlated D0s) to CLEO III data (uncorrelated D0s). The errors onthe predictions are propagated from the total errors on rK∗K and δK∗K . The errorson the ratios from data are statistical errors.
136
6 Conclusions
In an analysis using 9.0 fb−1 of data collected at√
s ≈ 10.58 GeV with the CLEO III
detector, we find the strong phase difference δD ≡ arg(
A(D0→K∗+K−)
A(D0→K∗+K−)
)= δK∗−K+ −
δK∗+K− = 332◦ ± 8◦ ± 11◦ and relative amplitude rD ≡∣∣∣A(D
0→K∗+K−)A(D0→K∗+K−)
∣∣∣ =aK∗−K+
aK∗+K−=
0.52 ± 0.05 ± 0.04. This measurement indicates significant destructive interference
between D0 → K+(K−π0)K∗− and D0 → K−(K+π0)K∗+ in the Dalitz plot region
where these two modes overlap.
The fit includes the K∗± and φ resonances and a non-resonant amplitude. The
measured fit fractions for each resonance (with statistical uncertainty only) are (46.1
± 3.1)% for the K*+, (12.3 ± 2.2)% for the K*−, (14.9 ± 1.6)% for the φ, and (36.0
± 3.7)% for the non-resonant contribution.
A second fit including the scalar κ± (mass 878 MeV/c2, width 499 MeV/c2) am-
plitudes instead of a non-resonant amplitude results in a strong phase difference of
δD = 313◦ ± 9◦ (stat.) and an amplitude ratio of rD = 0.52 ± 0.05 (stat.) The
measured fit fractions for each resonance (with statistical uncertainty only) are (48.1
± 4.5)% for the K*+, (12.9 ± 2.6)% for the K*−, (16.1 ± 1.9)% for the φ, (12.6 ±5.8)% for the κ+, and (11.1 ± 4.7)% for the κ−.
We also investigate the D0 → K+K−π0 Dalitz plot from 281 pb−1 of data collected
at√
s ≈ 3.77 GeV with the CLEO-c detector. We find that the Dalitz plot exhibits
the same behavior as the Dalitz plot seen in the CLEO III analysis, which is confirmed
by a fit to the data.
137
A The K-matrix Formalism
In two-body scattering of the type ab → cd, the cross section in the partial wave J
is [16]:
σJfi =
(4π
q2i
)(2J + 1)|T J
fi(s)|2, (A.1)
where ‘i’ and ‘f ’ stand for the initial and final states, s is the square of the center-of-
mass energy, qi is the breakup momentum (see Equation 2.5 for the definition) in the
initial system, and T Jfi is the transition amplitude from the initial state to the final
state. In the following sections, it is important to note that we suppress references to
the total angular momentum J . Therefore, a separate K-matrix is required for each
partial wave.
A.1 The K operator
The scattering operator, S, is the operator which gives the amplitude, Sfi, that an
initial state |i〉 will be found in the final state |f〉 [16]:
Sfi = 〈f |S|i〉. (A.2)
We may separate S into a part where the initial and final states do not interact (the
identity operator, I) and a part that defines the possible transitions, which we define
138
to be the transition operator T :
S = I + 2iT . (A.3)
The factors of 2 and i (here i refers to the imaginary number√−1) have been intro-
duced for convenience. Since probability must be conserved, the scattering operator
S must be unitary:
SS† = S†S = I . (A.4)
We can multiply each side of Equation A.3 by its conjugate transpose (on the left,
then separately on the right) and we find:
T − T † = 2iT †T = 2iT T † (A.5)
We may write this in terms of inverse operators with a small amount of algebraic
manipulation, and then transform it further:
(T †)−1 − T−1 = 2iI
(T−1 + iI)† = T−1 + iI (A.6)
This allows us to introduce a new hermitian operator, K defined as:
K−1 = T−1 + iI (A.7)
Then we may eliminate the inverse operators in Equation A.7 through multiplying
by K and T from left and right and vice versa:
T = K + iT K = K + iKT (A.8)
139
Notice that T and K commute.
Both T and S may now be expressed only in terms of K:
T = K(I − iK)−1
= (I − iK)−1K, (A.9)
and S = (I + iK)(I − iK)−1
= (I − iK)−1(I + iK) (A.10)
A.2 The K-matrix
Resonances appear as a sum of poles in the K-matrix, which contains the informa-
tion regarding how the K operator relates initial to final states. When resonances
dominate the scattering amplitudes we write the K-matrix elements as [16]:
Kij =∑
R
gRi(m)gRj(m)
m2R − m2
, (A.11)
where
g2Ri(m) = mRΓRi(m). (A.12)
The sum is over the number of resonances each with mass mR which decay to two-
body channels i and j, and ΓRi(m) is the partial width for the scattering of the initial
state through the resonance R. The partial widths are given by the expression:
ΓRi(m) =ΓRi(mR)
ρi(mR)[F
(J)Ri (m)]2ρi(m), (A.13)
140
where
ρi(m) =2pi
m, (A.14)
F(J)Ri is the normalized Blatt-Weisskopf barrier factor of Table 2.1 for the decay of
R (spin J) to decay channel i, and pi is the breakup momentum for the decay of
resonance R into decay channel i. It should be noted that ΓRi(mR) is precisely the
measured decay width of resonance R to channel i.
A.3 K-matrix Examples and Argand Plots
A.3.1 One Pole, One Decay Channel
For only one resonance which decays to a kaon and a π0 at pole mass mR, K is simply
written as:
KR→Kπ=mRΓR(mKπ)
m2R − m2
Kπ
(A.15)
We may determine the T -matrix which contains the information about the transition
from initial to final states from Equation A.9 and we find:
TR→Kπ =mRΓR(mKπ)
m2R − m2
Kπ − imRΓR(mKπ)(A.16)
As with the Breit-Wigner formalism, we expect the rate at a particular value of m2Kπ
to be proportional to the amplitude squared, |T |2. Also, we can plot the imaginary
part of T (Im[T ]) vs. the real part of T (Re[T ]) for all values of m2Kπ in order to
confirm that unitarity is conserved. We call this type of plot an Argand plot.
The Argand plot for the decay of the K∗(892) resonance to Kπ is shown in
Figure A.1. Note that the Argand plot is a circle of unit diameter with its center at
141
-0.4 -0.2 0.2 0.4
0.2
0.4
0.6
0.8
1 2 3 4 5
0.2
0.4
0.6
0.8
1
1.2
mKπ
2
|T|2
Re[T]
Im[T]
mKπ
2 = mK*(892)
2
Re[T] = 0
1
Figure A.1: (left) An Argand Plot of the K∗(892) T -matrix for all values of m2Kπ.
(right) A plot of |T |2 vs. m2Kπ.
{Re[T ] = 0, Im[T ] = 12}. This circle is called the unitarity circle, and the physically
allowed values of T should remain at or within this circle, in order to preserve the
unitarity of S.
The Argand plot for the decay of the K∗(1410) resonance to Kπ is shown in
Figure A.2. Even though the resonance has a different mass and width, the Argand
plot looks identical to the Argand plot for a lone K∗(892) resonance.
A.3.2 Two Poles, One Decay Channel
Naively, we could add the amplitudes (or T -matrices, in this case) for the K∗(1410)
and K∗(892) in order to determine the combined amplitude for the decays of the two
Kπ resonances, as we do in the Breit-Wigner formalism. We can see the Argand
plot for the sum of T -matrices in Figure A.3. It is very clear that unitarity is no
longer conserved, as the path of the Argand plot is not the circular shape we expect.
The Argand plot does go near {Re[T ] = 0, Im[T ] = 1}, but not at the peaks of
142
-0.4 -0.2 0.2 0.4
0.2
0.4
0.6
0.8
mKπ
2
Re[T]
Im[T]
mKπ
2 = mK*(1410)
2
Re[T] = 0
1 2 3 4 5
0.2
0.4
0.6
0.8
1
1.2
|T|2
1
Figure A.2: (left) An Argand Plot of the K∗(1410) T -matrix for all values of m2Kπ.
(right) A plot of |T |2 vs. m2Kπ.
mKπ
2
Re[T]
Im[T]Re[T] = 0 (not exactly at the peaks...)
-0.4 -0.2 0.2 0.4 0.6
0.2
0.4
0.6
0.8
1 2 3 4 5
0.2
0.4
0.6
0.8
1
1.2
T ≠ 0 (!)
1
|T|2
Figure A.3: (left) An Argand Plot of T = TK∗(892)+TK∗(1410) for all values of m2Kπ < 5.
(right) A plot of |TK∗(892) + TK∗(1410)|2 vs. m2Kπ.
143
mKπ
2
Re[T]
Im[T] Re[T] = 0
-0.4 -0.2 0.2 0.4
0.2
0.4
0.6
0.8
1
1 2 3 4 5
0.2
0.4
0.6
0.8
1
1.2
T = 0
|T|2
Figure A.4: (left) An Argand Plot of T calculated from KK∗(892) + KK∗(1410) for allvalues of m2
Kπ < 5. (right) A plot of |T |2 vs. m2Kπ.
the resonances. Also the Argand plot noticeably does not pass through {Re[T ] =
0, Im[T ] = 0} between the two resonances, as would be expected if we stayed on
the unitarity circle. The K∗(892) and K∗(1410) are separated in m2Kπ enough where
|T |2 is not affected too much, but for resonances which are close to each other the
preservation of unitarity decreases further.
We can preserve unitarity by adding the pole terms in the K-matrix and then
finding the associated T -matrix, instead of adding the T -matrices for the individual
resonances. For two resonances at different pole masses, both decaying to Kπ we can
define the K-matrix as the combination of the K-matrices for the scattering through
resonance R and resonance Q:
KKπ = KR→Kπ + KQ→Kπ, (A.17)
where the form of KQ→Kπ is the same as Equation A.15 (Qs replacing Rs).
144
The Argand plot of T when derived from the addition of K∗(892) and K∗(1410)
K-matrices is shown in Figure A.3. Unitarity is now preserved, and |T |2 becomes
zero between the two resonances.
A.4 Application to Dalitz Plot Analysis
A.4.1 Lorentz-Invariant T -Matrix
The T -matrix as we have defined it so far is not Lorentz invariant. The Lorentz
invariant T -matrix, which we call T , is given by [16]:
T = (ρ)−12 T (ρ)−
12 , (A.18)
where the “phase-space matrix” is a diagonal matrix which has:
ρii =2pi
m, (A.19)
where pi is the breakup momentum of channel i. All of the other elements of ρ are
zero.
The Lorentz invariant K-matrix, K, is given by:
K = (ρ)−12 K(ρ)−
12 , (A.20)
and then:
T = (I − iKρ)−1K, (A.21)
145
D0 P (1 - iK)-1a
b
c
possibledecay
channels
Final State
rescatteringProduction
Figure A.5: A diagram of the process where a resonance is produced from a D0 decay.The resonance may decay to a number of decay channels, if possible. The amplitudefor the decay channel with particles a and b can be found in the ab component of thefinal state vector �F .
A.4.2 The �P -Vector Formalism
Strictly, the K-matrix formalism applies only to s-channel resonances observed in the
two-body scattering of the type ab → cd. However, if we assume that the two-body
system in the final state is an isolated one and that the two particles do not interact
with the rest of the final state, we may generalize the K-matrix formalism to the
production of resonances.
We may define the production vector �P [17] as
Pi =∑
R
β0RgRi(m)
m2R − m2
, (A.22)
where we require the poles of the �P -vector to be the same as the pole of the K-matrix.
�P is a n-dimensional vector where n is the number of decay channels. The factors β0R
are complex amplitudes with units of energy which carry the production information
of the resonance R.
The production vector, �P , may be thought of as the virtual two-particle states
which the resonances decay to (see Figure A.5). These virtual states scatter to the
146
final state which is described by the final state vector, �F :
�F = (I − iK)−1 �P = TK−1 �P . (A.23)
A.4.3 Practical Application
In its Lorentz invariant form, the production vector is written as:
�P = ρ− 12 �P , (A.24)
and the final state vector is written as:
�F = ρ− 12 �F , (A.25)
so the invariant final state vector can be written simply as:
�F = T K−1 �P. (A.26)
One Pole, One Channel
From Equations A.15, A.22, and A.24 we can determine the �P -vector for a single
resonance decaying to the Kπ channel:
PKπ =β0
R
√mRΓR→Kπ(mKπ)
(m2R − m2
Kπ)√
ρKπ
. (A.27)
We can rescale β0R as follows:
βR =β0
R√mR
ΓR→Kπ(mR)ρKπ(mR)
(A.28)
147
and then use Equations A.26 and A.28 to get the �F -vector:
FKπ =βRmR
ΓR→Kπ(mR)ρKπ(mR)
m2R − m2
Kπ − imRΓR(mKπ)× F
(J)R→Kπ(m) (A.29)
Excluding the barrier factor, Equation A.29 is the K-matrix justification for the Breit-
Wigner model we define in Chapter 2, since the amplitude can always have arbitrary
normalization in a Dalitz plot analysis, so long as the probability distribution function
for the fit is properly normalized.
Two Poles, Two Channels
One of the motivations for thinking about the K-matrix formalism was the small dip
at m2Kπ ≈ 1 GeV2 /c4 in our data as seen in the m2
Kπ projections of Figure 4.11. Also
we wondered if the dip affected our measurements of rD and δD. Our concerns were
mostly resolved by changing the fitting axes from (m2K−π0 ,m2
K+π0) to (m2K+π0 ,m2
K+K−)
and by evaluating the systematic errors due to fitting different regions of the Dalitz
plot where the dip was not a factor. However, at the time we wanted to see how using
the K-matrix formalism with the �P -vector approximation could help us fit our data.
We believed that it was possible that the dip near m2Kπ ≈ 1 GeV2 /c4 may have
had to do with the interference of two J = 1 resonances in the Kπ decay channel. We
were certain that one of those resonances was the K∗(892) that we could clearly see,
and we thought the interfering resonance would be the next higher K∗ resonance, the
K∗(1410). We also thought that the preferred decay of the K∗(1410) to the K∗(892)π
channel might have the effect of causing a dip in the amplitude at K∗(892)π threshold
which is just above 1 GeV2 /c4.
To construct the �P -vector which represents the K∗(892) and K∗(1410) decays to
Kπ and K∗π, we should note that we may separate the �P -vector into two parts which
148
each represent the two resonances:
Pi = PK∗(892)i + PK∗(1410)i, (A.30)
where we define:
PRi =β0
R
√mRΓR→i(m)
(m2R − m2)
. (A.31)
for resonance R and decay channel i.
We choose the first �P -vector element to represent the Kπ decay channel, and
the second �P -vector element represents the K∗(892)π decay channel. The K∗(892)
resonance cannot decay to the K∗(892)π channel, so there is only one non-zero element
in the K∗(892) �P -vector:
�PK∗(892) =
⎛⎜⎝ β0
K∗(892)
√mK∗(892)ΓK∗(892)→Kπ(m)
0
⎞⎟⎠ × 1
(m2K∗(892) − m2)
. (A.32)
The K∗(1410) resonance can decay the both channels, so we can write the corre-
sponding �P -vector as:
�PK∗(1410) =
⎛⎜⎝ β0
K∗(1410)
√mK∗(1410)ΓK∗(1410)→Kπ(m)
β0K∗(1410)
√mK∗(1410)ΓK∗(1410)→K∗π(m)
⎞⎟⎠ × 1
(m2K∗(1410) − m2)
.(A.33)
The amplitude for these J = 1 resonances decaying to the Kπ channel is then
simply the first component of the corresponding �F -vector which comes from the total
�P -vector (�PK∗(892)+ �PK∗(1410)). The resulting amplitude is different for varying choices
of partial widths and βs. An example of the type of amplitude with a dip near 1
GeV is shown in Figure A.6. In the case of maximum destructive interference, the
relative complex phase between β0K∗(892) and β0
K∗(1410) is equal to zero, and the �F -
149
0.5 1 1.5 2 2.5 3
0.2
0.4
0.6
0.8
1
mKπ
2 (GeV2/c4)
|FK
π|2
Figure A.6: A plot of the magnitude of the �F -vector amplitude to the Kπ decaychannel squared vs. Kπ invariant mass squared for destructively interfering K∗ res-onances.
vector is exactly zero in between the resonances. The partial widths used are shown
in Table A.1. The magnitudes of the βs were arbitrarily set to make |FKπ|2 ≈ 1 at
the resonance peaks. In an actual Dalitz plot analysis, of course, the βs would be
free parameters in the fit.
We attempted a fit (to the Dalitz plot shown in Section 4.4) which used the
above P -vector formalism for the Kπ resonances, and also had a Breit-Wigner φ
Resonance (R) Decay Channel (i) Partial Width (ΓRi(mR))K∗(892) Kπ 52.7 MeV /c2
K∗(1410) Kπ 15.2 MeV /c2
K∗(1410) K∗π 190.9 MeV /c2
Table A.1: Partial widths used to calculate �F for interfering K∗ resonances.
150
resonance and a non-resonant contribution. The only differences in the P -vector
formalism we used for the fit are that we chose ΓK∗(1410)→Kπ = 15.0 MeV /c2 and
ΓK∗(1410)→K∗π = 200 MeV /c2. We fitted to the Dalitz plot variables (m2K−π0 ,m2
K+π0)
in this example, and the projections of the fit are shown in Figure A.7.
The dip could also be the result of interference between S-wave resonances. In
a future high-statistics Dalitz plot analysis of D0 → K+K−π0 it will be worth con-
sidering that the dip could possibly be an interference in the S-wave for the κ and
K∗0(1430) resonances, which could be fit with a K-matrix parametrization [56]. This
may possibly involve a large coupling of the K∗0(1430) to Kη′ (mη′ = 957.78 ± 0.14
MeV, Γη′ = 0.203 ± 0.016 MeV [1]). Another interpretation of the dip from Refer-
ence [56] is a possible preference for resonances which would decay to Kπ to decay
to Kη instead (mη = 547.51 ± 0.18 MeV, Γη = 1.30 ± 0.07 keV [1]). If this is the
case, there would be an enhancement at Kη threshold in the D0 → K+K−η Dalitz
plot. However, this interpretation is considered to be less likely since the coupling of
a J = 0, P = +1 resonance1 to Kη is expected to be suppressed [56].
1J = Total angular momentum, P = Parity eigenvalue
151
mK
+K-2 (G
eV2 /c
4 )m
K-π
02 (GeV
2 /c4 )
mK
+π
02 (GeV
2 /c4 )
Events per 0.03 GeV2/c
4
Fig
ure
A.7
:A
nex
ample
ofa
fit
usi
ng
the
P-v
ecto
rfo
rmal
ism
.
152
References
[1] W.-M. Yao et al. [Particle Data Group], Journal of Physics G 33, 1 (2006).
[2] F. Halzen and A. D. Martin, Quarks & Leptons: An Introductory Course inModern Particle Physics (John Wiley & Sons, New York, 1984).
[3] W. E. Burcham and M. Jobes, Nuclear and Particle Physics (Longman, Harlow,United Kingdom, 1995).
[4] C. Plager, Ph.D. Thesis, University of Illinois at Urbana-Champaign, 2003.
[5] D. Griffiths, Introduction to Elementary Particles, (John Wiley & Sons, NewYork, 1987)
[6] A. D. Sakharov, Sov. Phys. Uspekhi 34, 392 (1991).
[7] P. Harrrison et al., The BaBar Physics Book: Physics at an Asymmetric BFactory (SLAC Report 504). (Stanford University, Stanford, CA, 1998).
[8] M. Gronau and D. Wyler, Phys. Lett. B 265, 172 (1991).
[9] D. Atwood, I. Dunietz, and A .Soni, Phys. Rev. D 63, 036005 (2001).
[10] Y. Grossman, Z. Ligeti, and A. Soffer, Phys. Rev. D 67, 071301(R) (2003).
[11] A. Giri et al., Phys. Rev. D 68, 054018 (2003).
[12] A. Poluektov et al. [Belle Collaboration], Phys. Rev. D. 73, 112009 (2006).
[13] J. Charles et al. [CKMfitter Group], Eur. Phys. J. C 41, 1-131 (2005). Updatedresults and plots available at: http://ckmfitter.in2p3.fr
[14] S. Kopp et al. [CLEO Collaboration], Phys. Rev. D 63, 092001 (2001).
[15] See review on page 713 of W.-M. Yao et al. [Particle Data Group], Journal ofPhysics G 33, 1 (2006).
[16] S. U. Chung et al., Annalen der Physik 4, 404 (1995).
[17] I. J. R. Aitchison, Nucl. Phys. A 189, 417 (1972).
[18] J. Blatt and V. Weisskopf, Theoretical Nuclear Physics, (Dover, New York, 1987)
[19] P. L. Frabetti et al. [E687 Collaboration], Phys. Lett. B 331, 217 (1994).
153
[20] H. Pilkuhn, The Interactions of Hadrons, (North-Holland, Amsterdam, 1967)
[21] A. Weinstein, CLEO CBX 99-55 (unpublished).
[22] Y. Kubota et al. [CLEO Collaboration], Nucl. Instrum. Methods Phys. Res.,Sect. A 320, 66 (1992).
[23] T. S. Hill, Nucl. Instrum. Methods Phys. Res., Sect. A 418, 32 (1998).
[24] G. Viehhauser, Nucl. Instrum. Methods Phys. Res., Sect. A 462, 146 (2001).
[25] CLEO Collaboration, CLEO-c and CESR-c: A New Frontier of Weak and StrongInteractions, CLNS 01/1742, http://www.lns.cornell.edu/public/CLNS/ (2001).
[26] M. Artuso et al., Nucl. Instrum. Meth. Phys. Res., Sect A 554, 147 (2005).
[27] CLEO Collaboration, The CLEO III Detector: Design and Physics Goals, CLNS94/1277, http://www.lns.cornell.edu/public/CLNS/ (1994).
[28] ‘QQ - The CLEO Event Generator,’http://www.lns.cornell.edu/public/CLEO/soft/QQ, unpublished.
[29] R. Brun et al., Geant 3.21, CERN Program Library Long Writeup W5013 (1993),unpublished.
[30] J. L. Rosner and D. A. Suprun, Phys. Rev. D 68, 054010 (2003).
[31] D. M. Asner et al. [CLEO Collaboration], Phys. Rev. D 54, 4211 (1996).
[32] R. Ammar et al. [CLEO Collaboration], Phys. Rev. D 44, 3383 (1991).
[33] H. Albrecht et al. [ARGUS Collaboration], Z. Phys. C 46, 9 (1990).
[34] O. Tajima et al. [Belle Collaboration], Phys. Rev. Lett. 92, 101803 (2004).
[35] M. Bauer, B. Stech, and M. Wirbel, Z. Phys. C 34, 103 (1987).
[36] P. Bedaque, A. Das, and V. S. Mathur, Phys. Rev. D 49, 269 (1994).
[37] L. L. Chau and H. Y. Cheng, Phys. Rev. D 36, 137 (1987).
[38] K. Terasaki, Int. J. Mod. Phys. A 10, 3207 (1995).
[39] F. Buccella et al., Phys. Lett. B 379, 249 (1996).
[40] D. Cronin-Hennessy et al. [CLEO Collaboration], Phys. Rev. D 72, 031102(2005).
[41] H. Muramatsu et al. [CLEO Collaboration], Phys. Rev. Lett. 89, 251802 (2002).
[42] D. M. Asner et al. [CLEO Collaboration], Phys. Rev. D 70, 091101 (2004).
154
[43] P. Rubin et al. [CLEO Collaboration], Phys. Rev. Lett. 93, 111801 (2004).
[44] D. M. Asner et al. [CLEO Collaboration], Phys. Rev. D 72, 012001 (2005).
[45] M. Ablikim et al. [BES Collaboration], Phys. Lett. B 633, 681 (2006).
[46] S. Teige et al. [E852 Collaboration], Phys. Rev. D 59, 012001 (1998).
[47] M. Ablikim et al. [BES Collaboration], Phys. Lett. B 607, 243 (2005).
[48] S. Eidelman et al. [Particle Data Group], Phys. Lett. B 592, 1 (2004).
[49] F. James and M. Roos, Comp. Phys. Comm. 10, 343 (1975)
[50] H. Albrecht et al. [ARGUS Collaboration], Phys. Lett. B 308, 435 (1993).
[51] J. Wiss and R. Gardner, “Estimating systematic errors,” E687 memo: E687-94-030, http://web.hep.uiuc.edu/e687/memos/DALITZ SYS.PS (1994), unpub-lished.
[52] M. Artuso et al. [CLEO Collaboration], Phys. Rev. D 70, 112001 (2004).
[53] B. Aubert et.al. [BaBar Collaboration], Phys. Rev. D 74, 091102(R) (2006).
[54] C. W. Chiang, Z. Luo, and J. L. Rosner, Phys. Rev. D 67, 014001 (2003).
[55] D. Asner and W. Sun, Phys. Rev. D 73, 034024 (2006).
[56] J. L. Rosner, Phys. Rev. D 74, 076006 (2006).
155
Author’s Biography
Paras Naik was born in Elmhurst, Illinois on 4 May 1978 to Pradeep and Aruna Naik
of Wood Dale, Illinois. After almost two years his younger brother, Samar, joined the
family. When he was three years old he moved to Addison, Illinois, which became his
hometown. During his formative years, he attended Lincoln Elementary School and
Indian Trail Junior High School. He then proceeded to Addison Trail High School
and graduated at the top of his class. He entered the University of Illinois at Urbana-
Champaign in the fall of 1996. He received his Bachelor’s Degree in Engineering
Physics with Honors in January 2000. He continued at the University of Illinois and
received his Master’s Degree in Physics in May 2001 while being active in the Physics
Education Research Group. He pursued his Ph.D. in Physics by participating in the
High Energy Physics Group. During his time with the group he worked on the CLEO
experiment in Ithaca, NY, where he resided between August 2003 and July 2005.
Upon his return he successfully published a research paper entitled “Measurement of
Interfering K∗+K− and K∗−K+ Amplitudes in the Decay D0→K+K−π0”2 with the
assistance of the CLEO Collaboration. He will continue particle physics research as
a Postdoctoral Fellow with Carleton University.
2C. Cawlfield et al. (CLEO collaboration), Phys. Rev. D 74, 031108(R) (2006).
156