Measurement of the Nuclear Dependence and Momentum Transfer Dependence of Quasielastic (eel P) Scattering at Large Momentum Transfer by Naomi C. R Makins B.Sc., University of Alberta 1989) Submitted to the Department of Physics in partial fulfillment of the requirements for the degree of Doctor of Philosophy at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY September 1994 Massachusetts Institute of Technology 1994 Signature of Author ... - C7 1 - - Department of Physics July 19, 1994 f Certified by ............ ....... j .. Richard G. Milner Associate Professor, Department of Physics Thesis Supervisor Accepted by ............ MASSACHUSETTS INSTITUTE OFTFCHNOinry I(CT 141994 LIBRARIES ' George F. Koster Chairman, Departmental Graduate Committee Acknowledgments Summarizing the contributions of the many individuals who have made my graduate years such a formative experience is infinitely more daunting than summarizing the contributions of deadtimes and radiative corrections to my graduate data. First of all, I must thank my advisor, Prof. Richard Milner, for providing a learning and working environment which has been truly ideal. I recall going for lunch some time ago with several fellow graduate students; with us was a prospective new student who was visiting MIT (and whose presence provided us willing volunteers with a free meal in exchange for answering her questions). I remember rhapsodizing enthusiastically on the exceptional research environment at MIT, and enumerating some of the details of my own work group with which I was so delighted. I remember, too, the subsequent stares of astonishment from my colleagues. Little episodes like listening to their equally rhapsodic lists of grievances about their groups make one deeply grateful for having an advisor who is so thoroughly concerned for his students' well-being. One must never take for granted someone who conducts study sessions twice a week to prepare his students for the oral exams, who comes up with scholarship opportunities out of thin air, who directs his students from Day One towards a thesis topic, who is willing to drive in from Arlington at ten o'clock at night to read yet another thesis draft. Richard's outstanding research program has provided each of his students with exceptional projects of enviable quality. I have greatly appreciated his ability to give one the trust and space to work alone, while keeping a watchful eye on the direction and relevance of one's endeavors I have countless times benefitted from his keen and astute perception. He seems able to constantly keep the end of a project in sight, to see directly the essence in a morass of detail and confusion, and to have an utterly accurate sixth sense of the way things will turn out. So many times, he has pointed out the Forest to me, while I analysed and plotted only the Trees. I fervently hope that his intuition and ability to see with a wider eye is a contagious one. I have been touched by the generous hospitality of Richard and Eileen Milner, who have many times invited me to their home for sumptuous dinners. It 2 has been a great pleasure knowing Eileen, who has always had an encouraging word for me and is a fabulous story teller. I have been fortunate all around in the people I had to work with on NE18. From the outset of the experiment, Rolf Ent was my immediate superior in the chain of command and managed to teach a starry-eyed youngster with her head full of the ineffable beauties of field theories how to be an experimentalist. I have benefitted tremendously, by deed and by example, from his efficiency, his vast knowledge and intuition in nuclear physics, his friendship, his tireless assistance on everything from analysis to getting this thesis written, and his ability as a swimmer. Tom O'Neill and Wolfgang ("The Phantom") Lorenzon formed the West Coast half of the NE18 task force, and it has been a pleasure to know and ork with such talented, able, and congenial people. I have been working closely with Tom since the beginning of the experiment and I couldn't ask for a more skilled colleague, or one more willing to share and explain his work in deatil. I am also tremendously thankful that his codes are so easy to read. I am grateful, too, to Brad Fillippone for numerous, extremely helpful local and long-distance discussions, and for his strange and wondrous ability to understand exactly what you mean even when you don't. Henk Jan Bulten deserves special mention for his expert assistance on so many stages of this experiment; for his truly unmatched ability with spectrometer optics he will forever be dubbed Mr. Matrix in my glossary. Finally, let me include John Arrington and Eric Belz, and we have the entire membership of the Hell Week Crew. It was a delight to be part of a group of people who were willing to do what it takes, and to have a good time at it too. (All we need is a faded group photo and squadron insignia to complete the war buddies clich6). Janice Nelson, now an Operator Extraordinaire, was an undergraduate at MIT when she worked for our group, and I cannot express how much I missed her greased-lightning efficiency when she graduated. She claims to have returned The Brain, but I have my doubts as I can't find it anywhere. And Dave "C++" Wasson, responsible for our radiative corrections prescription, is the type of theorist that every experimentalist should 3 have at the end of a Red Phone direct line. He is a wizard, a joy to work with, brilliant and patient with his explanations, and I believe a highly skilled golf player. You couldn't ask for more perfect running conditions than what we experienced at SLAC. It was a rare treat to spend a year at a facility entirely staffed by helpful and efficient people who were so clearly happy with their work. "It must be the weather!", we exclaimed, but for whatever reason the MCC operators delivered a beam so stable that a Monte Carlo simulation could hardly have done better. The American University group must get special mention for their tireless and skilled assistance to us newcomers at the End Station. A heartfelt "thank you" and the hope that we can do it again sometime go to Lisa Andivahis, Ray Arnold, Peter Bosted, Thia Keppel, Allison Lung, Steve Rock, Linda Stuart, and Zen Szalata. Finally, no summary of the SLAC membership to whom a debt is owed would be complete without a mention of Mr. 0. Katt, whose vigorous efforts did so much to keep the experimenters on their toes. I would like to thank my thesis committee members, Prof. Ernie Moniz and Prof. John Tonry for their encouragement, valued comments, patience, and flexibility - what ordi- nary committee would consent to read a student's thesis during weekends, in unstapled segments hastily squeezed under office doors, and ultimately consent to attend a defense on a Sunday? I am particularly grateful to Prof. Moniz for allowing me to profit from his vast experience and understanding via discussions and astute observations. I am also grateful for his appreciation of Canadian dialects and figures of speech. Despite the weather, the MIT physics department is the most accessible, concerned, and thoroughly human example of a bureaucracy that I have ever encountered. I have been deeply impressed with the series of faculty-student roundtables and seminars on the employment issues facing young physicists which the department has sponsored. The willingness of the graduate student office to accomodate students by bending rules and deadlines to the breaking point is unprecedented in my experience. Here, I must express my deepest gratitude to Peggy Berkovitz, who has done so much for all of us and makes the graduate office a delightful place to visit. Peggy knows just exactly how grateful I am for 4 the aformentioned flexibility, and by some means unknown to modern science or ancient alchemy has managed to keep my affairs in order for five years despite all of my unwitting efforts to thwart her. I will also miss Joanne Gregory, who keeps life in order here at the Laboratory for Nuclear Science. We've had some great conversations and some good laughs; I'll ne,,,er forget the words of encouragement that she has provided at exactly the right moments. My classmates are a wonderful group of people, consciencious, thoughtful, and so many with widely varied interests and fascinating backgrounds. Studying for the general exams produces yet another set of war buddies, and I remember fondly all our discussions, study sessions, and the old Friday dinners. I know every one of them will go far. A special thanks to Jordina, Ole, Mike T, and Mike Y for some truly great times. I'm thankful that physics is a small world - I'm confident I'll see you all again soon, and often. Same thing goes to Bryon and Eric up at Bates - keep the tradition alive, you two! Thanks to Kevin, for patiently teaching me all about the polarized target when I first got here, and for helping me pack on the very last crazy day! To Thia, the kind of friend that comes along once in a very blue moon, you've always been there for me and I can only hope to do the same. Finally, I'm giving a copy of this thesis to Deborah Scott, and another one to Mark Davey, so that they know how much I enjoyed the late night rap sessions and pep talks. Good luck to both of you! To thank my family would be to thank a limb or a sense for its existence. What we do, we do together; the best of what I am is what they have given me. This work, as always, is theirs. 5 Measurement of the Nuclear Dependence and Momentum Transfer Dependence of Quasielastic (ee'p) Scattering at Large Momentum Transfer by Naomi. C. R. Makins Submitted to the Department of Physics on July 19, 1994, in partial fulfillment of the requirements for the degree of Doctor of Philosophy Abstract Experiment NE18, performed at SLAC, has measured the coincidence quasielastic crosssection for (ee'p) scattering from Q of I to 68 (GeV/c)'. This extends the existing Q' range of such measurements by over an order of magnitude. Five targets were used: 11, 2H, 12C, 56Fe, and 197Au. To test our understanding of quasielastic scattering, the data were compared with a Monte Carlo calculation of the experiment based on a conventional nuclear physics picture. This calculation included radiative effects, using a prescription based on the work of Mo and Tsai and recalculated in a coincidence framework. The elastic hydrogen data were found to be well explained by standard parametrizations of the proton form factor. Spectral functions were extracted from the nuclear data and found to be in good agreement with the Plane Wave Impulse Approximation (PWIA), the deForest offshell electron-proton cross section ,,,, and Independent Particle Shell Model spectral functions based on measurements made at Q _ 0.2 (GeV/C)2. The nuclear transparency was extracted from the data, and examined for evidence of colour transparency. This phenomenon, motivated by perturbative QCD considerations, is predicted to cause a rise of the transparency with Q2 . No evidence of such a rise was observed in the data. Also, the A-dependence of the transparency was found to be well parametrized by a classical model of transmission through the nucleus. Thesis Supervisor: Richard G. Milner Title: Associate Professor, Department of Physics 6 Contents 1 Introduction 15 1.1 The Plane Wave Impulse Approximation 16 1.2 The Off-Shell 1.3 Previous (ee'p) Cross-section Data 1.3.1 Independent 1.3.2 Spectroscopic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Particle Shell Model Sum Rule . . . . . . . . . . . . . . . . . 1.3.3 Evidence for Multi-body Processes . . . . . 1.3.4 Nuclear Transmission . . . . . . . . . . . . . 1.4 Exclusive Processes at High Momentum Transfer 1.5 of Perturbative 1.4.1 Application 1.4.2 Counting Rules and Other Results of PQCD 1.4.3 Comparison 1.4.4 Arguments in Favour of a Higher Perturbative, Threshold with Data Colour Transparency 1.5.1 Experimental QCD . . . . . . . ............ ............ ............ ............ ............ ............ ............ ............ ............ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 22 25 27 32 34 34 37 40 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Evidence 21 . . . . 2 Description of the Experiment 41 44 50 52 2.1 B eam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 2.2 Experimental 55 Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.3 Target .. . . . . . . . . . . . . . . . . . . . 2.4 1.6 GeV/c 2.5 8 GeV/c 2.6 Trigger Electronics 2.6.1 1.6 GeV/c 2.6.2 8 GeV/c and Coincidence Triggers Spectrometer . . . . . . . . . . Spectrometer 2.7 Data 2.8 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . Trigger Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................. ................. ................. ................. ................. ................. ................. ................. 3 Data Analysis 3.1 Overview 3.2 Tracking 56 58 66 70 70 71 73 74 79 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1 1.6 GeV/c Tracking 3.2.2 8 GeV/c Tracking 3.2.3 Multiple Tracks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Reconstruction of Events to Target . . . . . . 3.4 Determination . . . . . . . . . 3.5 Tim ing . . . . . . . . . . . . . . . . . . . . . . 3.6 Corrections of E, and p . . . . . . . . . . . . . . . . . 3.6.1 Electronic Deadtime Corrections 3.6.2 Computer Deadtime Corrections 3.6.3 Proton Absorption . . . . . . . . . . . .............. .............. .............. .............. .............. .............. .............. .............. .............. .............. . . . . . . . . . O verview . . . . . . . . . . . . . . . . . . . . . 4.2 Off-Shell Prescription for o, I I I I I I I I I I 8 79 80 83 84 86 89 90 93 93 95 104 108 4 Description of the Experimental Simulation 4.1 79 ............... ............... 109 110 4.3 . . . ill . . . 116 Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 Model Spectral 4.3.1 Functions Correlation 4.4 Spectrometer 4.5 Radiative . . . . . . . . . . . . . . . . . . . . . . Corrections Internal . . . . . . . . . . . . . . . . . . . . 4.5.1 First Order Bremsstrahlung 4.5.2 Virtual Photon 4.5.3 Higher Order 4.5.4 Peaking Approximations . . . . . . . . . . . . . . . . . . . . . . 141 4.5.5 External Brernsstrahlung . . . . . . . . . . . . . . . . . . . . . . 150 4.5.6 Radiative Techniques Employed in the PWIA Calculation . . . 152 Corrections Bremsstrahlung . . . . . . . . . . . . . . . . 125 . . . . . . . . . . . . . . . . . . . . . 131 . . . . . . . . . . . . . . . . . . . . 138 5 Results of the Experiment 5.1 Extraction of Results 154 . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Acceptance Cuts 5.1.2 Extraction of the Spectral 5.1.3 Extraction of Transparency 5.1.4 Systematic Uncertainties 154 . . . . . . . 155 . . . . . . . . . . . . . . . . . . . . 157 . . . . . . . . . . . . . . . . . . . . . . 157 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Measurement . . . . . . . 161 . . . . . . . 176 . . . . . . . 177 Hydrogen 5.3 Spectral Function 5.4 Nuclear Transparency Function . . . . . . . . . . . . . . . . 5.2 Results . . . . . . . . . . . . . . . . . . . 154 . . . . . . . . . . . . . . . Measurement with Glauber . . . . . . . . . . . . . 5.4.1 Comparison Calculations . . . . . . . 5.4.2 Comparison with Colour Transparency Predictions . . . . . . . 184 5.4.3 A Dependence . . . . . . . 188 . . . . . . . . . . . . . . . . . . . . . 6 Discussion of Results 195 9 List of Figures 1-1 PWIA 1-2 E, i-3 model of the (ee'p) projection of "C projection of 1-4 Energy dependence 1-5 Spectroscopic 1-6 Effect of correlations 1-7 Bates measurement reaction . . . . . . . . . . . . . . . . . . . . . 19 spectral function . . . . . . . . . . . . . . . . . . . 23 C spectral function . . . . . . . . . . . . . . . . . . . . 24 of the nucleon-nucleon . . . . . . . . . . 26 factor for various nuclei . . . . . . . . . . . . . . . . . . . . 27 . . . . . . . . . . . . . . . . 28 in the dip region . . . . . . . . . . . . . . 29 1-8 Saclay measurement of 2C(ee'p) in the quasielasticregion. . . . . . . . . 30 1-9 Bates measurement of 31 1-10 Nuclear transmission on spectroscopic of C(ee'p) RT at Q and RL -_ 034 factor cross-section structure functions from 2C(ee'p) . (GeV/c) . . . . . . . . . . . . . . . . . 33 1-11 Schematic diagram of factorization in PQCD analysis of electromagnetic form factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 1-12 Born diagrams contributing to a PQCD evaluation of the nucleon form factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-13 Example of Landshoff diagram in meson-baryon scattering 37 . . . . . . . . 39 scaling at high Q . . . . . . . . . . . . . . . . . . . . . . . . . 41 1-15 Proton-proton cross-section compared with PQCD prediction . . . . . . . 42 1-16 Elastic proton form factor G' 43 1-17 PQCD calculations of Fl'(Q'), using various proton wavefunctions . . . . 44 1-14 Form factor 10 Contribution of non-perturbative amplitudes to G' - - I . . . 45 1-19 Evidence for geometric scaling of hadron-hadron cross-sections 47 1-20 Colour transparency calculations et al . . . . . . . . . 49 1-21 (p,2p) data of Carroll et al . . . . . . . . . . . . . 51 1-18 transparency diagram of Farrar 2-1 Schernatic of the A-beamline . . . . . . . . . 54 2-2 Floor plan of End Station A . . . . . . . . . . . . . . 57 2-3 Schematic drawing of the 16 GeV/c spectrometer 2-4 Schematic drawing of the 16 GeV/c detector stack 2-5 Arrangement of wires in the drift chambers of te 2-6 Schematic drawing of the 2-7 Schematic diagram of the 2-8 Formation 2-9 Formation of the 61 . . . . . . . . . . . 16 GeV/c spectrometer 65 GeV/c spectrometer . . . 67 eV/c detector stack of the 16 GeV/c trigger 69 . . . . . . . . . . 72 Q 2== 1(GeV/c) GeV/c and coincidence triggers 73 2-10 Raw and corrected coincidence timing spectra for "'Au at Q of 68 (GeV /C)2 2-11 63 75 Experimental phase space in E, and p, for the carbon measurement at . 3-1 Timing 3-2 Example of raw coincidence timing spectrum measured by TDC8- 3-3 Proton 4-1 E,, and p 4-2 Feynman diagrams contributing to first order Bremsstrahlung cross-section 127 4-3 Feynman diagrams contributing to the virtual radiative correction . . . . 133 4-4 Angular 5-1 E windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .78 used in selection absorption as a function distributions distribution for of coincidence events . . . . . . . . . . 92 - - - - 101 of Q2 . . . . . . . . . . . . . . . . . . . . 106 H(ee'p) . . . . . . . . . . . . . . . . . . . . of first order Bremsstrahlung photons . . . . . . . . distribution of H(ee'p) events, compared with PWIA calculation 11 124 143 162 5-2 E, distribution of 'H(ee'p) events, compared with PWIA calculation 163 5-3 Extracted spectral function for 164 integrated over 2H, d3p, 5-4 Extracted spectral function for 12C, integrated over 5-5 Extracted spectral function for 5-6 Extracted spectral function for "'Au, integrated over d3p,,, 167 5-7 Extracted spectral function for 2H, integrated over dE, 168 5-8 Extracted spectral function for 12C, integrated over dE,,, 169 5-9 Extracted spectral function for 56Fe , integrated over dE, 170 integrated over 56Fe , 165 d3p,,, 166 d3p, 5-10 Extracted spectral function for "'Au, integrated over dE, 171 5-11 Extracted p(p,) for the 1p shell of 2C . . . . . . . . . . . 174 5-12 Extracted p(p,,) for the Is shell of 2C . . . . . . . . . . . 175 5-13 Measured nuclear . . . . . . . . . . . . . . . 178 transparencies 5-14 Measured transparency for '2C, compared with Glauber calculations . . . 180 5-15 Measured transparency for 16 Fe, compared with Glauber calculations 181 5-16 Measured transparency for "'Au, compared with Glauber calculations 182 5-17 Transparency 185 calculations of Benhar 5-18 Correlation effects on transparency, et al . . . . . . . . . . . . . . . . . . as calculated by Nikolaev et al. . . 186 5-19 Measured transparency for 12C, compared with colour transparency calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 5-20 Measured transparency for "Fe, compared with colour transparency calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 5-21 Measured transparency for "'Au, compared with colour transparency calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-22 Measured transparencies, vs. A, compared with best fit classical model 12 191 194 List of Tables 1.1 Transverse contribution to the free ep cross-section at NE18 kinematics. 32 2.1 Description 59 2.2 Description of materials in the path of incident and scattered particles 60 2.3 Summary 77 3.1 Correction 4.1 Minimum 4.2 Model spectral function 4.3 Model spectral 4.4 Model spectral 4.5 Correlation corrections applied to IPSM model spectral functions 4.6 Single photon Brernsstrahlung in the soft photon approximation, at Q = I (G eV / 4.7 of targets of kinematic factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . settings applied to the data . . . . . . . . . . . . . . . . . . . . 107 for several nuclei . . . . . . . . . . . 114 parameters for 12C . . . . . . . . . . . . . . . . . 116 function parameters for separation energies . . . . . . . . . . . . . . . . 117 function parameters for . . . . . . . . . . . . . . . 118 56Fe 197 C)2 proton Au . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 131 Single photon Bremsstrahlung cross-section in the soft photon approximation, for 100 MeV photon 4.8 . . . . . . . . . . . . . . . . . . . . . . . energy Radiative correction functions vided into various components . . . . . . . . . . . . . . . . . . . . . . 131 for single photon Bremsstrahlung, subdi. . . . . . . . . . . . . . . . . . . . . . . . 136 4.9 Singlephoton Brernsstrahlungin the ultra-relativistic limit . . . . . . . . 137 13 5.1 E, and p cuts placed on results 5.2 Cuts placed on reconstructed spectrometers quantities 5.3 Summary 5.4 Summary of the measured nuclear transparencies 5.5 Input values used in classical model of transmission . . . . . . . . . . 193 5.6 Fit values of . . . . . . . . . . 193 of systematic . . . . . . . . . . . . . . . . . . . . . . uncertainties . . . . . . . . . . 156 . . . . . . . . . . . . . . . . . . . . 160 . . . . . . . . . . . . . 177 ff using classical model of transmission 14 155 Chapter Introduction Quasielastic electron scattering from nuclei provides an excellent means of studying nuclear structure. One notable advantage to the use of electrons is that the interaction is relatively weak, and so an electron probe is able to sample the entire nuclear volume. Another is that the electron-photon interaction vertex is extremely well understood in terms of QED A wealth of quasielastic (ee'p) data exists on a host of nuclei, at Q up to about 03 iGeV/c)'. (Here, Q denotes the square of the four-momentum transferred to the struck proton.) These measurements have been interpreted with considerable success in terms of the Plane Wave Impulse Approximation (PWIA) and the Independent Particle Shell Model (IPSM). In reference to this latter model, (ee'p) measurements have provided conclusive evidence for the existence of a shell structure in the nucleus, and have been able to measure precisely the properties of individual shells. Several phenomena requiring further explanation have also been identified, such as violations of the spectroscopic sum rule and evidence of contributions from multiparticle currents. If we take (ee'p) measurements to higher Q, it is likely that, as the wavelength of the scattering probe becomes smaller and individual nucleons are more precisely singled out, the PWIA description will improve. On the other hand, it is possible that such an 15 approach will cease to be valid because, at energies of the same magnitude as the proton mass, one might expect quark degrees of freedom to begin to play a role and eventually dominate the reaction dynamics. At sufficiently high energies also, the asymptotic freedom of the strong interaction should allow perturbative methods to come into play and permit the exact calculation of QCD processes. The threshold energy at which PQCD becomes a valid approach is at present under debate. A recent prediction of PQCD known as "colour transparency" provides us with a signature that can be effectively sought using coincidence electron scattering. Experiment NE18 was performed to investigate all of these questions by taking (ee'p) data at Q values up to 68 (CeV/c) - an order of magnitude higher than previously achieved. 1.1 The Plane Wave Impulse Approximation Elastic electron scattering from a free nucleon in lowest order QED (i.e. one-photon exchange) is described by the Rosenbluth cross-section: where Iq 12 _ Q2 and do, do, dQ dQ Mott I = 1 1+2(1+-r)tan2 I [GE' +,rE-'G M 2] This expression involves the Mott cross-section 2 (for a spinless relativistic electron scattering from a fixed point charge), and two form factors GE and GM which describe the electric and magnetic structure of the nucleon. For reference, these form factors can also be expressed in the Dirac-Pauli form: KQ2 GE F - 4M2F2 Gm F + F2, 16 (1.2) where is the anomalous magnetic moment of the nucleon 1-79 for the proton and 1.91 for the neutron). A further reprentation is in terms of the transverse and longitudinal response functions: RT 2 = 27-GM RL =(I )G2E' (1-3) These form factors have been measured up to momentum transfers of (GeV/c)' for the proton and 4 (GeV /C)2 for the neutron [1]. (Neglecting the small contribution of GP E at very high Q2 , GP M has in principle been measured up to Q2 = 31 (GeV/C)2 [21). However, when one places the nucleon in a nucleus and attempts to describe quasielastic scattering in a similar way, one is immediately faced with several fundamental difficulties. First, the nucleon is bound and so is off-shell; the relevance of free form factors to this situation is not clear. Also, the nuclear electromagnetic current which couples to the virtual photon now depends on the dynamical structure of the nucleus: it is not simply that of an isolated charge/current distribution, but depends on the complex interactions of the nucleon with its surroundings. Determining a complete, coherent solution to this problem requires a thorough understanding of the nuclear wavefunction and should be calculated using a relativistic theory. This task is sufficiently formidable that it has not yet been accomplished for even the simplest nucleus, 2H. In practice, one adopts an approximate approach known as the impulse approximation (IA), wherein the nucleons are treated as independent entities moving in a mean potential. This approach, though straightforward, leads at once to an inconsistency. The current conservation relation d dt s7 J = -P 17 (1.4) required of any isolated system can only be satisfied if one includes exchange currents between the moving charged particles[3]. Such currents are by definition neglected by the impulse approximation. Nevertheless, the IA has historically proven remarkably successful in describing (ee'p) data. A second approximation helps to simplify the description. This is the Plane Wave Approximation, which assumes that the scattering matrix elements can be evaluated using free unperturbed particle wave functions (i.e. plane waves). Also assumed is that one-photon exchange is sufficient to describe the ep scattering vertex (the Born approximation). The validity of each of these assumptions should improve as the reaction energy increases. First, the Feynman amplitude for virtual photon emission is proportional to , where q. is the -momentum of the emitted photon. The relative strength of higher order to first order diagrams thus decreases with increasing momentum transfer and the one-photon exchange approximation improves. Second, high energy particles are less affected by either the electromagnetic or strong fields of the residual nucleus than particles of lower energy. The plane wave assumption of using unperturbed particle wavefunctions should therefore improve as well. The assumptions described above constitute the Plane Wave Impulse Approximation, or PWIA A schematic diagram of this model of the (ee'p) reaction is presented in Figure 1-1. The variables used throughout the text to describe the reaction kinematics are as follows: le,k] = 4-vector of incident electron IC/,k/] = 4-vector of scattered electron [E, p] = 4-vector of initial state proton (bound) [E PI] [ErecPreel 4-vector of scattered proton = 4-vector of recoiling A-1 system 18 --a. [E", Pl [ C, k] --M. E rec Prec I Figure 1-1: Schematic diagram of the PWIA model of the (ee'p) reaction. 1W,q] q., 4-vector of exchanged virtual photon Q 2 -qjjqt = q q - w2 IVIP proton mass MA target mass Note that it is only by virtue of the plane wave assumption that we can define a momentum for the iitial state proton at all. Using all the information available in a coincidence measurement and assuming this model of the reaction, we can deduce the initial state proton 4-vector from the "missing" energy and momentum (i.e. the 4-vector missing from the energy-momentum conservation relation): EM = E C - (E - Mp - E, - Mr), (1-5) pm = (1-6) p'- q. 19 Since the PWIA is assumed throughout our analysis, these variables will often be used interchangeably with E and p. Note that the recoiling A-1 system is not observed. Its kinetic energy is evaluated by observing that momentum conservation in this picture requires Pree :-- -pn,, and by assuming that M, = MA-,. This last relation is not strictly true, as the recoiling system may be in an excited state and have a mass larger than its ground state mass. However this variation makes only a small contribution to E,, one that cannot be detected without very high energy resolution. The components of the vector pm are generally defined relative to the momentum transfer q and the scattering plane. They are referred to as "parallel" (pm 4), "perpendicular" "out-of-plane" (pm Wx k). (pm-Wx Wx 0)), and As will be explained in Section 28, NE18 took measurements in so-called "perpendicular kinematics". This refers to the fact that the large values of p, detected by the experiment had large perpendicular components. A sign was therefore attached to the variable p, corresponding to the sign of pm, , as defined above. Note that positive p, indicates missing momentum vectors on the larger-scattering-angle side of the q vector. A direct consequence of the PWIA model is the factorization of the differential doublearm cross-section into nuclear structure information and bound eN cross-section 4: d6o, dE'dQdE'dQp K is a kinematical factor, equal to Pp'. = KO',NS(Ep). The spectral function S(Ep) (1.7) is interpreted as the probability density of finding a nucleon with momentum p and energy E in the nucleus. It is defined to be S(p, E) < 1la+b(E' +H P 20 - - Ei)ap It > Here, H is the nuclear Hamiltonian, E is the energy of the initial nucleus in eigenstate Iand a and a are the usual particle creation and annihilation operators. 0,N is the fundamental off-shell electron-nucleon cross-section, and is discussed in the next section. 1.2 The Off-Shell Cross-section The problems involved in determining a consistent model of quasielastic scattering have been addressed by many workers. One of the prescriptions most commonly used in that of deForest 3], who imposes the current conservation requirement by a suitable choice of gauge field. He expresses the off-shell cross-section N in terms of four nucleon structure functions: O'eN O'Mott A2w C + A - 2 tan 2(-O) WT 2 A A tan 2(_10) 2 ,Q--7 = arccos(k k, IqlI WI Cos 2 + where A 1/2 and ACOS2+ tan 2(i 0) Ws 2 arccos[(4 x k) (4 x p)]. expressions for the structure functions are given in ref (1.9) Two different 3 based on two different off- shell extrapolations of the nucleon current. These are named and O',,2", where the subscript, "cc" refers to the current conserving nature of the forms; both versions were used in our calculations. The structure functions are expressed in terms of the nucleon form factors, and another assumption involved in this prescription is that free form factors can be used. 21 1.3 Previous (ee'p) Data 1.3.1 Independent Particle Shell Model Coincidence (ee'p) experiments to date have been performed at Q up to about 03 (GeV/c)', and have produced a wealth of information about the nucleus. In particular, dramatic evidence of the shell structure of the nucleus has been provided. Figure 1-2 depicts the results of two measurements of the missing energy distribution for 12C. In panel (b), one can clearly see the two shells expected in the ground state, with the 1S112 peaking at zero momentum, and the IP3/2 dominating in the higher momentum slice. Panel (a) presents results from a higher precision experiment performed at Saclay, with missing energy resolution of I MeV. This precision reveals additional structure in the spectrum, corresponding to the various possible excitation states of the residual nucleus. These states affect E, via the term T, - B the kinetic energy of the recoiling A-1 system which depends on the mass MA-,. Figure 13 shows the p, spectrum for carbon as measured by three different experiments. One can isolate p(p"') for individual shells by placing missing energy cuts around the appropriate peak. The solid line in the figure does an excellent job of describing the distribution, and is the result of solving for the nucleon wavefunction in a Woods-Saxon mean field potential. Such an approach is referred to as the Independent Particle Shell Model (IPSM). To produce such agreement, however, it is necessary to evaluate the distortions of the scattered proton wavefunctions caused by interactions with the spectator nucleons. Corrections for these effects constitute the Distorted Wave Impulse Approximation model, or DWIA. An Optical Model (OM) potential is used to compute the proton distortions, containing both a real term which shifts the measured particle vectors relative to their true vertex values, and an imaginary term which represents the possibility of inelastic scattering in the final state 22 I IQC. CP) (4 h 1 1 C4 E U en a- I "i I U 1 O VI W 2 A U -10 W"M 0V W M W W EWRGY(MM a) b) Figure 12: E distribution for carbon, Figure 76 from ref 4. and causes the "absorption" (or loss) of scattered protons. A further enhancement of DWIA involves the evaluation of electron distortions as well, calculated in the Coulomb field of the nucleus. These calculations are referred to as CDWIA (Complete DWIA), and are especially important for heavy nuclei. It is important to note that the need for these complex corrections should decrease as one proceeds to higher energies. Clearly, a higher energy electron will be less affected by a fixed change in electric potential. As for the proton, Figure 14 shows the energy dependence of the NN interaction and reveals that the inelastic portion of the cross-section dominates at nucleon momenta greater that about 23 GeV/c. Consequently, at such en- IVW -1 w ap 4L P. 8 - (Mew/0 Pe Figure 13: P,,, distribution for carbon, Figure 91 from ref. [4]. 24 MMICI ergies the proton final state interactions will be largely absorptive - the proton will be lost rather than deflected. This simplifies things greatly, making it a good approximation to correct for final state interactions (FSI) with a single factor. 1.3.2 Spectroscopic Sum Rule The theoretical distributions in Figure 13 have been normalized to agree with the magnitude of the data, and so say nothing about the total amount of strength actually found in the extracted spectral function. From the IPSM, one expects that the integral of the spectral function over a shell with total angular momentum eracy (2j+l) - will equal the shell degen- i.e. the number of protons in the shell. This relation is often referred to as the spectroscopic sum rule. In fact, one finds empirically that the integral has a lower value. The left panel of Figure 1-5 shows the strength found in the valence orbitals of several nuclei relative to the sum rule expectation. Only about 70% of the expected strength is found. Attempts to explain this depletion focus on the nucleon-nucleon correlations that are neglected by a mean field theory. These many-body effects can be grouped into two broad groups. Weaker (sometimes called "long range") correlations shift nucleons to deeper binding energies, causing for example filling of the "empty" orbitals above the Fermi energy (see the right-hand panel of Figure 1-5). "Short range" correlations on the other hand are caused by the strongly repulsive nature of the nucleon-nucleon force at short separations. An interaction with such a strongly correlated pair will greatly increase the measured missing momentum, shifting strength to well beyond the Fermi momentum kF. These correlations will therefore cause an overall decrease in the spectral function below kF- alculations of correlation effects are frequently done using nuclear matter models (an approximation to the nuclear medium using infinite boundary conditions). 25 400 II I. I. II II 11II I I .I .I .I I. II I. II I, ..,,1 1 .I I I . II I I. .I .I .I .I -I I I I .I . II- I I I I I 11 102 O'total 'Z E ?Iio if b - _ 41, -4 44- . j t .#ff4 -- - a ti I444t 4 I f # - . . 0 . I I f , , t -, 4 I 10 I-_ r- J I I I I I, I b 10'I - I I I , I I I 11 100 1.9 (7eaafic , f; + I?, I f f I II I 102 I 101 I 3 2 Pbeam (GeV/0 I ! I I I I I I . I . 10. 6 6 7 I 4 I I I I , , , ,I I I I I I 103 I 20 104 I I I . 40 50 I 30 . . I I . . . I1001 ECM (GeV) 2000 ... I .I -. I I . . , I . I I . I . I''I I I I I . 11 I I I I I IIII I .1 V t- '3 -2 ,= lu (7total 7 b -41 f_ztr + f I f I 10 atotal I 1 f I I I I I , I , I I 10' fII--"t*- - -- , I , 1. 6elwtic I I . I , .,I 100 I 2 I 2.9 3 I I It '++t I . I , , ,,I 102 10I (GeV/c) Pbeam 1.9 + .1 f j 4 3 1 4 5 I I 5 EC. T_ 6 7 I I __ 7 I __1 910 20t 910 I I I 20 30 30 40I I 1 40 50 60 (GeV) Figure 14: Energy dependence of the nucleon-nucleon cross-section (ref. [5]). 26 U.0 t 'F=MPTV'r)PA1T.q (.8 1 - 0.6 0.6 - 0.4 0.4 z (n 30si A 0.2 (.2 12c 160 (D nn ID lo, 102 TARGET MASS I . ,,.I 40,48Ca9oZr (D D (D I lo, . . CD . . _(DI o, b. Figure 1-5: Evaluation of the spectroscopic factor for the valence orbitals and "emtpy" orbitals (above the Fermi energy) of various nuclei (ref. 61). Such models have been quite successful in explaining the depletion of valence orbitals (see for example Figure 16). Note that correlation effects do not appear to alter significantly the form of the p distribution below the Fermi momentum, a evidenced by results such as those of Figure 13 showing excellent agreement with mean field distributions. 1.3.3 Evidence for Multi-body Processes Inclusive (ee') experiments such as that of ref. 7 have indicated evidence of processes as yet unaccounted for in quasielastic scattering. Measurements in the so-called "dip" region (between the quasielastic peak and first nucleon resonance) have produced cross-sections larger than those predicted by the combined effects of quasielastic nucleon knockout, resonant A production, meson-exchage currents, and pion production [8]. Also, measurements of the transverse and longitudinal structure functions in the quasielastic region have revealed R RL ratios which are -_ 60% larger than the predictions of the impulse approximation with free nucleons [9]. One calculation 7 of scattering in the dip region 27 E 0) C P V; a) Y I- 9 fA M 0 0 -EM - F [MeV] P. Figure 16: Spectroscopic factor for valence orbitals of various nuclei, accompanied by predictions of correlated nuclear matter spectral functions (dotted line). The solid line includes surface effect corrections for 'O'Pb (ref. 61). manages to explain some of the missing strength by including the probability of scattering from multi-nucleon currents. These currents are caused largely by short-range correlations, a direct consequence of the strongly repulsive nature of the nucleon-nucleon interaction at short distances. This strong interaction makes it possible for an incoming electron to scatter off a coupled system composed of two or more quarks. Physically, such a process might be expected to knock out both nucleons rather than just one. The net effect would be the addition of the second nucleon's momentum and kinetic energy to the measured missing momentum and energy. These quantities can be measured directly in an exclusive reaction, and such studies have been performed. Ref. [101details a measurement of the (ee'p) reaction from `C in the dip region; the experiment was performed at Bates at Q = 012 (GeV/c)'. The resulting missing energy distribution is shown in Figure 17, and clearly shows an almost constant distribution 28 500 400 Ir i 210 300 L -.5 0 z ,% 200 0.0 P C.) 100 n 20 20 60 100 140 180 MissingEnergy (MeV) Figure 17: Missing energy distribution for Bates at Q = of strength at E 12 C(ee'p) in the dip region, measured at (GeV/C)2. up to 150 MeV (the maximum value detected by the experiment). A similar measurement performed at Saclay 111on "C in the quasielastic region, however, measured up to 80 MeV in E, at Q2 = 0 16 (GeV/C)2 and detected no evidence of strength unexplained by single-particle knockout (see Figure 1-8). It is possible, of course, that the relatively small contribution from quasielastic scattering in the dip region enables one to detect the presence of additional effects. The measurement of the transverse to longitudinal structure function ratio has also been investigated with coincidence experiments. "C(ee'p) Ref. performed at Bates which separated the RT 121 details a measurement of and RL structure functions at quasielastic kinematics. Figure 19 depicts the extracted structure functions, measured out to > 60 MeV in missing energy at Q2 = 0 14 (GeV/C)2. The differenceST - SL is shown in panel (c); here, STandSLrefer to the spectral function of equation 17, but with 29 M15SING ENERGY MeV) Figure 1-8: Missing energy distribution for "C(ee'p) sured at Saclay at Q = 06 (GeV/c)'. in the quasielastic region, mea- only the transverse or longitudinal term of o,p included respectively. The IPSM, PWIA formalism makes no provision for any difference in these two versions of the spectral function. The data indicate that ST and SL are about equal at the valence 1p shell, but that the transverse portion dominates increasingly as one moves to higher missing momentum and eventually the difference appears to level off. As pointed out in ref. 12], since the density varies from shell to shell one might suppose that excess strength seen in the ls but not the 1p shell is indicative of a density-dependent modification of the bound nucleon current. However, the fact that this extra strength appears only in the transverse reaction channel suggests the presence of an additional process. If indeed these measurements are indications of coupling to correlated multi-nucleon currents, it is of great interest to repeat them at high momentum transfer. First, the cross-section is predominantly transverse at high Q. Table 1.1 shows the percentage transverse contribution to the free ep cross-section at the momentum transfers measured by NE18: above 80% at the lowest Q and above 98% at the higher settings. This should enhance our sensitivity to multinucleon processes. However, any cross-section for 30 1.00 (a) x- 3 0.75 0.50 0.25 0.00 2.0 IT .7.7 - (b) 3 - - - - - - - - - - 1.5 1.0 0.5 0.0 V 1 > -- - - - - 30 7 (C) 44 2-Body Thresh. 7 20 7 X lo - - - - - - - - - - - - 10 Figure 19: Measurement RL) from "J(ee'p) at Q2 transverse and longitudinal depicts RL, and (c) depicts 20 30 40 50 60 Missing Energy (MeV) of transverse and longitudinal structure functions (RT and = 04 (GeV/C)2, performed at Bates. ST and SL are the spectral functions, defined in the text. (a) depicts RT, (b) the difference ST - S. 31 Table 1.1: Percentage of the free ep cross-section which is transverse, at NE18 kinematics. I Q I transverse 1.04 3.06 5.00 6.77 contribution (%) ] 82.6 98.6 99.5 99.7 scattering from a composite object necessarily drops with increasing momentum transfer: the wavelength of the scattering probe decreases until it is able to resolve the substructure of the ob.ect, and at this point the large scale structure of the object ceases to play a role. It is thus reasonable to suppose that at Q values of the same order as the nucleon mass, the coupling of the virtual photon to correlated nucleon pairs is substantially suppressed. 1.3.4 Nuclear 9ransmission Final state interactions were mentioned above, and it was indicated that knockout protons are not only deflected but also absorbed by the spectator nucleons. Processes of this type are caused by the cross-section for inelastic final state interactions, and can be expected to increase with energy until the inelastic NN cross-section reaches its asymptotic value of about 28 mb. The quantity typically studied in this context is the "nuclear transmission", the probability of escape of a quasielastically scattered proton. Figure 1-10 shows transmission data for a variety of nuclei taken at Bates at Q' 0.34 (GeV/c)'. The transmission in this experiment was determined using the ratio of the coincidence (ee'p) rate to the rate of electron singles. A series of calculations are also depicted in the figure. The basic theoretical technique involved is that of Glauber [13], consisting of a semi-classical calculation of proton multiple scattering in the nuclear medium. The theory curves show the cumulative effect of incorporating various subtleties 32 z0 05 Ln 2 z Cn Cr. I- NUCLEON NUMBER Figure 1-10: Measurement of nuclear transmission at Q2 = 034 (GeV/c)2 14]. The lines refer to a calculation using the free Np cross-section (dotted), the adding Paull blocking (dashed), density-dependent effects (dot-dashed), and a correlation hole (solid). into the calculation. These are the Pauli blocking of protons scattered into filled states below the Fermi momentum, the density-dependence of the nucleon-nucleon cross-section in the nuclear medium, and the inclusion of two-body short range correlations. The full calculation appears to account successfully for the measured final state losses. These and other elements of FSI calculations, along with their importance at higher energies, will be discussed in detail in Section 54. 1. 33 1.4 Exclusive Processes at High Momentum Transfer Dramatic evidence of the existence of sub-nucleonic degrees of freedom was provided by the celebrated parton experiment of 1967. The observation of Bjorken x-scaling above Q2 of I (GeV/c)' (and for missing mass W > 2 GeV) in deep inelastic electron-proton scattering indicated the presence of point scatterers inside the nucleon.With the fact thus established that inclusive electron scattering at relatively low Q2 can resolve this parton substructure, one is led to wonder to what extent quark kinematics contribute to nuclear structure and to the dynamics of quasielastic scattering. Exclusive processes at large momentum transfer test both the internal dynamics of hadrons and the detailed structure of hadronic wavefunctions at short distances. Electron scattering at Q2 > (GeV/C)2, for example, corresponds to a spatial resolution of less than about 0.1 frn and so is clearly able to distinguish between the component partons of a nucleon. Calculating the properties of the nucleus directly from CD, however, is a formidable task. The strong coupling constant a, diminishes with increasing energy, and is larger than I until one reaches the region of so-called asymptotic freedom. At sufficiently high energies, a, becomes small enough that one can apply perturbative methods, of the type that have proved so successful in QED. This technique is referred to as PQCD. The energy threshold at which PQCD becomes a viable description, however, is not clear 15][24]. In the attempt to identify this threshold, the measurement of exclusive processes plays a dominant role. 1.4.1 Application of Perturbati've CD An excellent presentation of the application of PQCD to exclusive processes is found in ref. 16]. As an example of the manner of analysis employed, we consider the calculation of the nucleon magnetic form factor GM. Treating the problem in the familiar infinite 34 X1 0-4 O Figure 111: Schematic diagram describing the factorization of PQCD scattering amplitudes into initial and final quark distribution amplitudes and O*,and a hard scattering amplitude TH. This diagram specifically describes the calculation of an electrornagentic form factor. momentum frame, the momenta of the quarks are taken to be parallel to the nucleon's direction of motion and are parametrized in terms of the Bjorken scaling variable x. This variable denotes the fraction of the nucleon momentum carried by each quark that F_ x (so 1). This assumption is equivalent to requiring that the average transverse momentum of the quarks, < k' > 2 is much less than Q, the only momentum scale in the problem. The nucleon is described in terms of a Fock state expansion over its valence and excited states, where the valence state consists of the minimal number of quark fields (qqq) and the excited states contain additional elements (qqqg, qqqqq, ...). The reaction itself is then factorized into three pieces, as is depicted in Figure 1-11: the incoming and outgoing nucleon wavefunctions O(xi, Q) and O*(yi, Q), and the hard scattering amplitude TH(xi, yi, Q). The wavefunctions (or "quark distribution amplitudes") are assumed to be in their 3-quark valence states, an approximation which is supported below. TH is the amplitude for the initial nucleon to absorb the momentum q of the virtual photon, and redistribute it among the component quarks to produce a final state where the quark momenta are also approximately parallel. This redistribution is required in an exclusive reaction: both the initial and final hadrons are measured, and so the quarks are bound in known configurations. The "spectator quarks" in the reaction must thus "catch up" to the struck ark so that the nucleon retains its identity without multiparticle emission. 35 To leading order, TH is calculated by summing the connected Born diagrams shown in Figure 1-12, which depict the incoming virtual photon and the gluon exchange described above. We see now why in the limit of high momentum transfer only the minimal Fock state of the nucleon contributes: the number of gluon lines increases with the number of elementary fields present, and according to the Feynman rules each of these propagators contributes a factor of Q-' to the amplitude. One explicitly takes the nucleon wavefunction, then, to be an integral of the 3-quark wavefunction Ov(xj, k i)over transverse quark momenta less than Q: O(xi, Q = I d'k-Lid2kl2d 2k.L30V(xi, k -)0(k2 -Ls I < Q2). (1.10) This restriction is equivalent to ensuring that the transverse spatial separation (or "quark impact separation") b-L is at least of order Q-1 - so that the virtual photon will resolve individual quarks. The final result for the form factor is Q2) GM( (Q2 Q2 2 E anm Q 2 -Yn -Ym log A2 1 + 0(a,(Q)) + 0 nm The logarithmic terms can be considered corrections to the leading order behaviour; they are parametrized in terms of the "anomalous dimensions" depend on the nucleon wavefunction used. a, Q2 (Q2) Q 2 Am, An scaling which is the running coupling constant of and also depends logarithmically on Q2. a, (Q2) In Q2 A2 36 (1.12) (a) (c) W I V I I I C9) d) (F)I I I I I I a i I I (q) I X Figure 112: Leading order diagrams contributing to a PCD form factor. 1.4.2 evaluation of the nucleon Counting Rules and Other Results of PCD In the mid-1970's, Brodsky and Farrar 171demonstrated that in the asymptotic limit of high momentum transfer, application of dimensional analysis to a renormalizable field theory such as QCD leads to a set of so-called "counting rules" for the energy-dependence of fixed-angle electromagnetic and hadronic scattering. The derivation of the counting rules is most straightforward for exclusive reactions, and only these will be discussed below. The analysis is based on the following assumptions: First, one postulates that a composite hadron can be replaced by point-like constituents carrying finite fractions of the hadronic momentum. This, of course, is the essence of the quark model and of PQCD, and an implicit assumption here is that the number of constituent fields in mesons and baryons is accurately given by the quark model to be 2 and 3 respectively. Next, it is assumed that the only scales (dimensional quantities) in the system are particle masses and momenta. Further, the asymptotic limit of high center-of-mass energy s is taken (thus the masses can be neglected), and all invariants are fixed relative to s (in a two- 37 particle reaction, this corresponds to fixing the center-of-mass angle). The result of this is that only one scale, s, remains in the problem. The crux of this assumption is that binding effects do not contribute any additional scales - i.e. that the scaling behaviour of bound quarks is well-described by that of a collection of free quarks. With these two assumptions, the leading-order (Born) diagrams involved in a perturbative evaluation of the scattering amplitude can be written down. One now performs dimensional analysis on the scattering amplitude M. From the Feynman rules, one knows that dimensionally, each external fermion line contributes [length] (via the normalization of the Dirac spinors), each photon or gluon propagator (internal line) contributes [length]-', and each fermion propagator contributes [length]-'. With some reflection one sees that the net dimension of the amplitude is entirely described by the minimum number n of elementary fields (i.e. leptons, photons, and quarks) in both the inital and final states: [M] = [length]n-4 . Given that Vs is the only length scale, one obtains the counting rule for two-particle scattering: da dt AB-CD := dom.tt Im dt 2 ,,82-n f (S), -t t I (S --+00,S fixed). (1.13) This can be generalized to the case of multiparticle production (see ref. 171). A corollary is the counting rule for electromagentic form factors of hadrons: FH(t) , tl-nff, where nH is the minimum number of elementary fields in the hadron. Note that this agrees with the dominant Q2 dependence of Equation 1.11 for the nucleon form factor. A third assumption is being made somewhat implicitly here, which is that the class of so-called Landshoff diagrams [18] can be neglected. An example of such a diagram for 38 11 0V Figure 113: The left panel depicts a Landshoff diagram for meson-baryon scattering, while the right panel shows two characteristic connected Born diagrams for the same process. The off-shell quark propagators are indicated by dots. meson-baryon scattering is shown in the left panel of Figure 1-13(a), while the right panel depicts typical Born terms for this reaction. The Landshoff diagram is characterized by a relative absence of off-shell quarks (i.e. of internal quark lines), and physically corresponds to the independent, elastic, on-shell scattering of pairs of constituents from different hadrons such that the final momenta are properly aligned. These diagrams have a slower fall-off with Q than those described before, but are presumed to be strongly suppressed -- for example by the sheer number (tens of thousands) of connected Born diagrams involved in a purely hadronic PQCD calculation. Fortunately this complication is not present in lepton-hadron scattering. Other PCD predictions of the dynamics of exclusive reactions exist, notably those concerning helicity conservation. An example of this is the prediction that non-helicity conserving amplitudes are suppressed relative to helicity conserving ones. A case in pc is the proton form factor 2 which is not helicity conserving, unlike the form factor Fl' which describes a quark spin flip transition. The PQCD expectation is that F,(Q2) F2(Q2) (1.15) Q2 at high Q2 . These ideas can be further explored through the use of polarized beam and target experiments, and the measurement of spin-dependent asymmetries 19]. 39 1.4.3 Comparisonwith Data How well do these predictions fare when compared with experiment? The success of the counting rules in predicting the momentum-transfer dependence of many elastic form factors is demonstrated in Figure 1-14. These form factors appear to achieve the expected scaling behaviour at Q of around 5 (GeV/c)'. The free proton-proton cross-section is shown in Figure 1-15, and is seen to exhibit the s-" falloff predicted by the counting rules. The fluctuations of the cross-section around the leading-order PQCD form appear to be explained by the model shown as a solid line in the figure. This is the heavy-quark threshold model of ref. 20], where the contribution of scattering from hadronic resonances at certain kinematics is included. Unlike a hard-scattering reaction, a reonance couples to the large-scale structure of the proton, hence the deviation from the simple PQCD scaling picture. The predicted falloff of 2(Q') relative to FI(Q2) was verified by the recent experiment NE11 [1], which separated the form factors for Q2 up to 8 (GeV/c)' for the proton and 4 GeV/c)' for the neutron. The signature looked for is that the ratio G& GM approaches a constant; this can be deduced from equations 12 and 1.15. Another notable success is the calculation of the proton form factor G' it is compared in Figure 1-16 with the data of ref 2 The gradual drop in the calculation above Q2 -_ 7 GeV/c)' is caused by the appearance of the running coupling constant a,(Q2) in Equation 111. One must point out, however, that these predictions are arbitrarily normalized. The normalization (and sign) of the calculation is strongly dependent on the choice of quark distribution amplitude for the proton and so provides a measure of this distribution (assuming that the theory can be applied validly at all). This sensitivity as well as the fact that certain choices of wavefunction are able to obtain the correct normalization are demonstrated in Figure 117. The conclusion of these successful comparisons would seem to be that PCD is a viable theory of exclusive processes at momentum transfers of 40 100 lo-, ib" C U_ lo-, TC CY ,a 10-2 10-2 10-3 10-4 0 2 4 C)2 6 (GeV2) Figure 114:: The measured elastic form factors of several quark systems; the form factors are mltiplied by the power of Q predicted by dimensional scaling, and appear to approach this scaling behaviour as Q increases. (GeV/c)' and. higher. 1.4.4 Arguments in Favour of a Higher Perturbative Threshold From one point of view, it is somewhat surprising that the counting rules are so successful in these kinematic regimes. An essential assumption in their derivation is that all invariants in the reaction are sufficiently large that the particle masses can be neglected, and a momentum transfer squared of (GeV/c)' corresponds to only times the mass of the proton. A more serious argument in support of a far higher threshold for the 41 10-2 N a) O 1-1 .0 10-3 61E 10-4 0 a I'D 10-5 6 8 P lab 10 12 14 (GeV/c) Figure 1-15: Measurement of the proton-proton cross-section at 90' in the center-ofmass 21]. The dotted curve is the - s-10 PQCD prediction; the solid line is the prediction of ref. 22] including the effects of strange- and charmed-particle production thresholds. validity of perturbative methods is presented by Isgur and Llewellyn-Smith 241. These authors demonstrate that calculations involving only "soft" non-perturbative amplitudes can achieve a magnitude of the same order as the proton from factor data (Figure 118). Their argument centers on the selection of proton wavefunctions, which was shown previously to be crucial in the normalization of PQCD calculations. Equ. 1.10 indicated that in the calculation of the scattering amplitude M, transverse quark momenta k are restricted to values less than Q. In other words, the amplitude is determined by soft, low k_Lportions of the wavefunction. Their calculations suggest that a quark distribution with < k 2 >I=2 300 MeV produces a value for GM which is two orders of magnitude below the data. Much larger values of this mean transverse quark momentum are possible, but using them then violates the Q2 > < k 2 > requirement on which the validity of the perturbative techniques rests. Physically, one can picture this as follows: Essential to the perturbative approach is the hard gluon exchanges required to redistribute the momentum transfer Q among the quarks. If, however, the nucleon wavefunction has sig42 I r - \.,.%.j r---" V I I r,%--- '_ I I I M- I 0.5 W (D 0.4 i 0. 0.2 0 a IT 03 -0 0 0.2 8 - --IIII 0 --I --- I 10 (2 20 30 PGOV/62] Figure 116: Measurement of the elastic proton form factor GI at high momentum transfer (ref. 21). Note that the (small) contribution of GE was neglected in the extraction of the result. nificant strength at k_Lof the same order of magnitude as Q, this redistribution may not be required since the struck quark may be accompanied by other quarks already moving along its redirected path. Furthermore, Isgur and Llewellyn-Smith find that wave functions with < k' > 2 = 300 MeV can generate soft non-leading which are as large as the data. This fact alone brings the application of perturbative techniques into question. It is clear that the energy threshold at which PCD yet to be answered definitively. 43 can be applied is a question that has 1-5 P-V 0 11-1 V> 1.0 1-1 0 CZ CF. a 0.5 tv -0 0 a 0 0 M a TnsiAs Tnionval -S. - '-0.3(GeV/c")2 I 10 Q" I I 20 30 ((GeV/c)2] Figure 117: PQCD calculations of Fl'(Q'), using various proton wavefunctions[23]. The predictions use distribution amplitudes of Chernyak and Zhitnitsky (CZ), King and Sachrajda (KS), and Gari and Stefanis (GS). 1.5 Colour 'Transparency About 10 years ago, Mueller and Brodsky suggested that at sufficiently high momentum transfer, the final (and initial) state interactions of hadrons with the nuclear medium in exclusive processes should be reduced, leading to the phenomenon termed "colour transparency". The argument is based on three assumptions 251. First, a hard (high momentum transfer) elastic scattering vertex involving a hadron should "select" a particle configuration of reduced size. In a PQCD picture of the reaction, the hard virtual photons/gluons which carry the transferred momentum scatter directly from the quarks. Also, since exclusive processes are being considered, one can demand that the scattering is elastic. Thus if a hadron is involved, it must remain in one piece despite the fact that momentum is delivered to its component parts independently. The momentum imparted to the hadron must be then be distributed among its component partons, and 44 a4 Gm Q G' (( M (GeV4) 0. W (Gew) Figure 1-18, Demonstration that in certain calculations [241,contributions from soft, non-perturbative amplitudes to GP (Q2) (dotted lines) are found to be of the same order M as the data. 45 this is accomplished through the exchange of gluons. Each of these gluons will carry on average an equal fraction of the total momentum; since this is high, the uncertainty principle demands that the distance the gluons have to travel be correspondingly small - of order Q in fact. In other words, the hadron state selected at the hard scattering vertex is compressed in size. The object of reduced size is often referred to as a point-like configuration (PLC). Another way of looking at this is that an accelerating colour current causes gluon radiation and so multiparticle final states. To exclude this possibility, one must postulate that the struck particle is point-like and therefore colour neutral. The second assumption, of "colour screening", is directly related to this last argument. One postulates simply that small particles have small cross-sections. This is the QCD analogue of the QED Chudakov effect, which predicts that ee- pairs have a small separation at the production vertex and consequently a reduced interaction with the surrounding medium. This effect has been observed experimentally 261; in QCD, evidence for such an effect is provided by the geometric scaling of hadron-hadron cross-sections: the total crosssection Oh,,h, is proportional to < r2>< r2 hi h2 > (see Figure 119). The third assumption is that the distance over which a PLC expands to its dressed (free) size is at least as large as the nuclear radius. Suppose the expansion time in the nucleon rest frame is to. The expansion time in the nuclear frame is then -ft = M Eto, and so becomes larger with increasing energy. At sufficiently high energy, and making the previous two assumptions, the hadron's interactions with the nuclear medium are certain to be reduced over a substantial fraction of its exit path, and will ultimately disappear leading to complete C4colour transparency" An example of one of the early calculations of the size of this effect is depicted in Figure 120 from ref. 28]. The final state interactions are modelled using an effective 46 1-9 .0 E I-j 0 40 30 20 10 0 Figure 119: Evidence for geometric scaling of hadron-hadron cross-sections 271. hadron-nucleon cross-section: eff O'hN W 01hN tot Here, 1h z - Ih T + < n 2 k2 > t t z I T - ((1h - Z) + O(Z - h)- Ih (1.16) denotes the full expansion distance of the hadron, z the hadron's distance from the interaction point, () the usual step function, and verse area of the PLC. Both 1h which can assume several values; a measure of the trans- OhN < 2 k2> t and z are measured in the lab frame. gives no colour transparency, is a parameter produces the "quantum diffusion" model, and 2 the "naive parton" model. The equation assumes a geometric interaction -- i.e. that the cross-section is proportional to the transverse area of the hadron. That area, in turn, is assumed to expand at the rate xt2 ,,z T. The "naive par- ton" model assumes that the quarks separate at the velocity of light, giving xt - to Similarly, the total expansion distance in this model is proportional to the radius of the dressed hadron times the time dilation factor: h Yv'-rL -f M. The "quantum dif- fusion" model provides a second model of the expansion. As in any diffusive process, the 47 rate of expansion is given instead by x2 , z (consider, for example, the rms displacement of a particle moving at fixed velocity through a medium and executing a random walk in the transverse direction). Accompanying this picture is an estimate to for the expansion time. Here, Eh < E, -I Eh > refers to the energy of the free hadron, and E, to the energy of the intermediate (excited) hadron state produced at the vertex. This value of to corresponds to the typical quantum oscillation period of a wavefunction consisting of a superposition of two energy eigenstates. As is seen in Figure 120, these models produce widely different estimates of the magnitude of the effect: from a 20 to 40% increase in transparency over the to 68 (GeV/c)' range of NE18. It should be noted that although colour transparency was originally postulated from PQCD arguments, recent work has demonstrated that PCD is not necessary for the existence of the effect. For example, one can show 291 that the emission of mesons from a baryon in a PLC is suppressed by - ( rPc Y. I'meson The interaction of the PLC with the medium is thus reduced in the same geometric way as described earlier, but in a conventional meson-exchange picture of the nucleus. Colour transparency thus provides us with a clear signature to look for in quasielastic (eelp) scattering: As the momentum transfer of the reaction increases, evidence for CT effects will appear as a drop in FSI and so a rise in transparency relative to the Glauber calculations described earlier. Around proton momenta of I GeV/c, the free NN crosssection (cf. Figure 14) approaches an asymptotic value of 40 mb. At these energies, then, Glauber calculations may be expected to be Q independent, leaving a clear CT signal. 48 1.0 0.8 A 0.6 Z a 0.4 0.2 0.0 10 20 30 Plab (Geft) Figure 120: Colour transparency predictions from ref. 281, using Equation 116. The dashed line is for = ("quantum diffusion" model), and the dotted line for = 2 ("naive parton" model). The = results for no CT is constant with energy and equal to the value of the curves in the Pab --+ 0 limit. 49 1.5.1 Experimental Evidence One experiment already performed expressly to look for colour transparency effects was that of ref. 30]. This was a p,2p) experiment performed at the Brookhaven National Laboratory. Note that the transparency found in (p,2p) scattering cannot be directly compared to (ee'p) results because in the former case, strong final state interactions are present in three channels rather than one. The BNL data was taken at incident proton momenta from 6 to 12 GeV/c and at a center-of-mass scattering angle of 90', providing a Q range of 48 to 10.4 (GeV/c)'. Five nuclear targets were used, (along with a CH2 target for calibration), and T was defined to be the cross-section ratio T (daldt)(p - p quasielastic in nucleus) A(duldt)(p - p elastic in hydrogen) (1.17) The data are shown in Figure 121 and exhibit two striking features: a clear rise in transparency up to Q2 = 8.5 (GeV/C)2 and a subsequent dramatic fall-off. Though it is tempting to interpret the initial rise as evidence for CT effects, the fall-off is clearly not explained by this picture and must be understood first. Various attempts to explain this behaviour have been made. Ref. 311observes that the fluctuations in transparency are correlated with the oscillations of the free proton-proton cross-secton around the S_1' dependence predicted by PQCD (see Figure 1-15). Dividing the quasielastic crosssection by the free pp cross-section as in Equation 117 is liable to produce oscillations in transparency, especially when one considers the fact that only one of the scattered proton vectors was completely measured in the BNL experiment. The second proton was detected in a non-magnetic spectrometer, and so the initial proton momentum (and consequently the effective center of mass energy s at the vertex) had to be deduced using the assurntion E = . The interpretation of the (p,2p) data is thus somewhat unclear, and the data have yet to be satisfactorily explained 25]. 50 1.00 F] T T I r I I I I I I f _TT_ t IIII- (a) Various Nuclei >_1 C) C: 0.50 + -Li 0 -C I- - a) (t 0. X 0.20 V) C: It -Al C 0.10 E__ A 0.05 III I 1 4 .11, )r CU -Pb I I I I I I I I I I I I I I I I I I I I I I IIII 0.5 U C 0) co 0. W r. (t 0.2 E_ 0.1 0 2.5 5 7.5 10 12.5 15 Incident Momentum Gev/c Figure 121: (p,2p) transparency data from ref. [301. 51 Chapter 2 Description of the Experiment 2.1 Beam The electron beam was generated by the Nuclear Physics Injector (NPI) at the Stanford Linear Accelerator Center (SLAC). The NPI became operational in 1985, and uses the last six klystron sectors of the SLAC linear accelerator to deliver high intensity electron beams of up to about 5.8 GeV to End Station A. The beam is pulsed, with a pulse length of 16 Ms and rate of 120 pulses/sec, yielding a duty factor of 2xlO-'. The number of electrons delivered in each pulse (or spill) varied from 2x1O'Oto NW', providing average beam currents from 03 to 9.5,vA. The SLAC energy doubler (SLED) was not used during the experiment. The experiment was performed in End Station A ESA), located just beyond the end of the linear accelerator and between the arms of the SLAC Linear Collider (SLC). The beam is directed to the End Station from the switchyard via the A-beamline (see Figure 2-1). The first section of the A-line is known as the A-bend and consists primarily of dipoles (B 10 17) which defined the beam energy. Focusing was provided by 4 quadrupoles (Q10-13), and the energy spread was determined by the slits S10 and SUL At a setting 52 of ±0.5%, over 90% of the beam passes through the slits. At the lowest energies, a setting Of 0.1 w used to limit the spread and aximize the energy resolution; at higher energies where the reaction rate was lower, the slits were opened to ±0.2% in order to increase the beam current and maintain an acceptable count rate. The beam spread consequently dominated the missing energy resolution at the higher Q settings. Beyond the A-bend, two pairs of horizontal and vertical steering coils (A 10 13) provide fine-tuning of the beam position at the target and can be adjusted by the experimenters. Roller screens RSI and RS2 consist of ZnS-coated plastic plates and can be moved in and out of the beam line by remote control from the counting house. Cameras directed at the screens allow the experimenters to view the beam position between runs since these screens fluoresce visibly when struck by the incoming electrons. Monitoring of the beam position and width during data taking was accomplished using another series of detectors, read out online by a DEC Microvax 11(referred to as the BCS, or Beam Control System). The first of these devices was a wire array positioned 20 m upstream of the target. The array consists of two planes of thin Al wires strung 06 mm apart; in one plane the wires are horizontal, in the other, vertical. Electrons striking the wires caused secondary eission that was measured by a series of ADCs (one per wire). The signals from the array thus provided a picture of the beam profile. Second, a pair of resonant microwave cavities located further upstream monitored the offset of the beam from the cavity center. Because of the distance between the wire array and the cavities, a measure of the beam angle was also obtained. The BCS provided continuous online monitoring of these detectors, and through a feedback loop was able to fine tune the steering coils. The position of the beam at the target was kept to within 2 mm of center and the beam angle to less than .05 mr. The beam shape at the target was an ellipse roughly 1 mm wide and 2 mm tall. 53 A-Bond TOMWO RS2 Tawts , A13 Al I \ 012 Cl 0 : co PMI PUS a.s. slo GI I 914 - 1 13 lo I MM will$ C12 1 12- 4 C C0111mawr PM PUnd ap a Oim*UPOIO I 0 a 1 ffe" SL Sk A S" FIB o ingot mm I Figure 21: Schematic diagram of the A-beamline 54 The quality of the beam was also monitored using two scintillators placed in the vicinity of the -target. The "good spill" scintillator was located approximately 0 m from the target, where it could monitor beam-induced scattering from the target. Its signal was displayed on an oscilloscope as a function of time, and so its shape provided a measure of the distribution of charge in each pulse. The beam tune was adjusted to keep this distribution as uniform as possible. The "bad spill" scintillator was placed near the beam pipe and slightly upstream of the target. It was shielded against scattering from the target, and so the signal in this scintillator indicated the amount of halo around the beam. The beam tune was adjusted to keep the bad spill level to a minimum. The charge in each beam pulse was measured by two toroidal coils surrounding the beam pipe, forming a transformer circuit with the beam itself as the primary winding. The measured charge was read and integrated over each run by the BCS. Also, a toroid calibration was performed before each run by sending a pulse of known charge down a wire running through the coils and adjusting the gain parameter in the BCS software to provide the correct reading. The uncertainty in the charge measurement was determined to be 0.5 b experiment NES performed previously at SLAC 32]. 2.2 Experimental Layout Figure 22 shows a floorplan of End Station A. The beam enters the hall from the west, intersects the target at the indicated pivot point, and continues on to a beam dump outside the End Station. Three spectrometers are shown in the figure. The 16 GeV/c was used to detect scattered electrons, and the GeV/c measured scattered protons in coincidence; the 20 GeV/c was not used in NE18. The spectrometers are positioned on circular rails and mounted on motorized support structures, and their angles with respect to te beam line can be adjusted by remote control from the counting house. 55 The spectrometer angle position is measured by an encoder which is read out in the counting house. The encoder determines the angle by keeping track of the cumulative rotation of the carriage drive gears. Surveys of the spectromters were done both before and after the experiment, and the difference between surveyed and encoder values for their angular position was found to be less than 0.01' for the 16 GeV/c and 002' for the GeV/c. The point at which the magnetic axes of the spectrometers intersects the beam line was also measured, for a range spectrometer angles. The intersection points varied by about 0.1 cm with angle, and the 16 GeV/c centerline was found to point 1.09 cm upstream of the pivot on average. No such offset was found for the GeV/c. 2.3 Target The target consisted of two parallel vertical ladders, holding the cryogenic liquid targets and solid targets respectively. The selection of targets wa's accomplished by moving the ladders remotely from the counting house. The ladders were enclosed in a vacuum sealed aluminum chamber, which imposed thin aluminum windows between the targets and the incoming and outgoing particles. The liquid target ladder held two 'H and two 2 H target cells, of 4 cm and 1 cm length respectively. (The long hydrogen target was not used in NE18). The cells themselves consisted of aluminum cylinders 32 cm in radius with rounded endcaps, and were constructed from Coors beer can blanks to provide a minimal and constant thickness of material between the enclosed liquid and the detectors. The liquid 'H and 2H was circulated between the targets and a cryogenic cooler at a flow rate of 2m/s. Platinum resistors and vapor pressure bulbs were located near the inlets and outlets of the target cells to monitor the target temperature and pressure. These values were very stable and were used to determine constant target densities for the experiment. The uncertainty 56 --I I i I I I \ Figure 22: Floor plan of End Station A 57 in the target thickness due to the combined uncertainties in density and target length (the latter due to shrinkage of the target upon cooling) was estimated to be < 0.7%. Also on the liquid target ladder were two aluminum "dummy" targets, consisting of solid aluminum pieces positioned to simulate the endcaps of the long and short liquid targets. The thickness of the dummy targets however was 95 times that of the empty liquid cells. Measurements of the scattering rate from the dummy targets were reduced by this factor in the analysis and subtracted from the liquid target results. The purity of the liquid 'H and 'H was measured after the experiment and found to be 99.94% and 99-68% respectively. The solid target ladder provided two targets of different thickness for each nucleus studied. The purity of these targets was found to be effectively 100%. The uncertainty in the thickness of the targets was estimated at 02%. Details of all targets used are presented in Table 21. The total amount of material encountered by the incoming electron and scattered particles is summarized in Table 2.2. The materials in this table cause the particles to lose or gain energy between the point where they are measured and the scattering vertex. Corrections for these energy shifts are made in the analysis. 2.4 16 GeV/c Spectrometer The 16 GeV/c spectrometer, shown in Figure 23, is a 90' vertical bend spectrometer consisting of a single dipole magnet surrounding a vacuum-sealed chamber and topped by a shielded detector hut. The vacuum chamber is sealed at either end by mylar windows (12 mil and 14 mil thickness for the entrance and exit foils respectively). The opening angle is limited by a fixed slit made of 12 inch thick lead and positioned 1959 m from the target pivot. The slit is flared in the dispersive (vertical) direction, with a front opening 58 Table 21: Description of targets. The radiation lengths for the solid targets are from the formula of Tsai (ref. 33]). For the liquid targets, the values come from the Particle Data Booklet, where a correction for molecular binding is made. Name Material Density Rad. Length Thickness (g/Cm') 4cm-LH2 4cm-LD2 15cm-LD2 C-2% C_6% Fe-6% Fe-12% Au-6% Au-12% 1H 2H 2H 12C 12C 56Fe 56Fe 197 197 Au Au (g/CM2) (CM) (g/CM2) (% rad. len.) 61.28 122.6 122.6 42-66 42.66 13.88 13.88 6.46 6.46 4.03 4.03 15.75 0.4097 1.1730 0.1064 0.2098 0.0206 0.04013 0.2840 0.6853 2.6782 0.8985 2.5724 0.8330 1.6425 0.3946 0.7688 0.46 0.56 2.19 2.11 6.03 6.00 11.83 6.11 11.90 0.0705 0.1701 0.1701 2.193 2.193 7.829 7.829 19.157 19.157 of ±9.07 cm vertically and ±9.30 cm horizontally. Thinner movable slits just before the fixed slit were also available but were kept open during all but a few checkout runs. The acceptance of the spectrometer was determined to be 340 msr (for a point target). The 1.6 GeV/c is described in detail in ref. 341. The field of the dipole magnet was measured before each data taking run by remotely inserting an NMR probe into the center of the magnet. The central momentum was then determined using the following formula: k'(central = 0075 * (B where 326) (2.1) is te magnetic field in Gauss and k' is in MeV/c. The maximum momentum that could b detected with the 16 GeV/c spectrometer was 147 GeV/c. The optical properties of the 16 eV/c are defined by the single dipole magnet. Slanted and urved pole faces introduce quadrupole-type fringing fields, producing pointto-point focusing in the dispersive (vertical) direction and line-to-point in the non- 59 Table 22: Description of materials interposed in the paths of particles (between the scattering vertex and the points at which they are measured). The superscript "L" indicates materials only relevant to liquid targets. Item Material Density Thickness (g/CMI) (CM) (g/CM2) (% rad. len.) seen by incoming electron wire arrays target entrance foil liquid target upstream endcap L Al Al Al 2.70 2.70 2.70 0.0041 0.0025 0.0076 0.0111 0.0068 0.0206 0.046 0.028 0.086 Al 2.70 0.0122 0.0329 0.137 Al mylar 2.70 1.39 0.0127 0.0064 0.0343 0.0088 0.143 0.022 Al mylar 2.70 1.39 0.0127 0.0356 0.0343 0.0494 0.143 0.124 Al Al 2.70 2.70 0.0305 0.0254 0.0824 0.0686 0.343 0.286 I seen by both scattered particles liquid target downstream endcap L liquid target side wallL liquid target side wall insulation L seen by scattered electron 1.6 GeV/c target exit foil 1.6 GeV/c entrance foil seen by scattered proton 8 GeV/c target exit foil 8 GeV/c entrance foil 60 FOCAL PLANE 6 I - -M*-i-,-KAM LINE - I I I r-----j I Figure 23: Side view of the 16 GeV/c spectrometer 61 I dispersive (horizontal) direction. "Point-to-point" indicates that particles with the same momentum are focused to the same position at the focal plane; "line-to-point" refers to focusing based instead on a particle's angle in a given plane. The magnet is designed so that the focal planes in both dispersive and non-dispersive directions are coincident. The spectrometer had a nominal momentum bite of ±5%. Larger values of could be detected, but TRANSPORT calculations performed before the experiment determined that the acceptance ceased to be flat (i.e. independent of particle angle) outside this range. Consequently, ±5% cuts in were consistently applied to the data. Details of the optical models used in the analysis are given in Sections 33 and 44. The detector stack used in NE18 is shown in Figure 24. It consisted of four detection systems: a Cerenkov counter and shower counter calorimeter for particle identification, a series of drift chambers for precise tracking, and scintillator planes for fast timing. Both the scintillators and calorimeter were segmented and were used in track selection. In the figures and text below, Cartesian coordinates x, y, z are used to describe positions in the stack. These variables are defined using the TRANSPORT convention: the x axis runs in the dispersive direction toward higher momentum values (thus down at the spectrometer entrance), z is the magnetic axis, and y is defined so as to provide a right-handed coordinate system. Also, refers to the percentage deviation of a particle's momentum from the central momentum of the spectrometer. The positions of all detector elements were precisely measured during a post-experimental survey. The rate of pions entering the 16 GeV/c spectrometer can reach 00 times the electron rate at the highest Q value measured by NE18, and so it is critical that pions be rejected efficiently at the trigger level. To make this particle identification, one can use the fact that highly relativistic particles travelling faster than the speed of light in a refractive medium emit Cerenkov radiation. Since the speed of light in a medium of refractive index 62 SHIELOING j I . 4 SHOWERCOUNTERS P8 PA XUYU CXCY1 UPPER SCINTILLATORS V 1 zI zI zI zI zI /I /I zI 41 41 ........................... WIRE - CHAMBERS BXBy I -......................... AxAy I ........................... XDYD WWER SCINTILLATORS 16 1 1 1 1 5 5 Z 1 rn C02 Cherenkov Counter 0 Figure 24: Diagram of the 16 GeV/c detector stack. The detector positions are to scale, but their thicknesses are in general exaggerated. 63 n is /n, one obtains a threshold momentum below which a particle of mass m will not radiate Cerenkov light: p(threshold) M - n1- (2.2) I As this momentum is higher for a pion than an electron, pions and electrons of the same momentum can be distinguished by selecting a material with a value of n that will cause only the electrons to emit Cerenkov radiation. The radiation is emitted in a characteristic cone of half-angle 0 = os-'(1/nO). (2.3) The Cerenkov detector used in the 16 detection stack consisted of an aluminum cylinder 1.3 m in height, with aluminum entrance and exit windows 04 mm thick, and filled with C02 gas of refractive index 100045 (at room temperature). A concave mirror of very thin aluminized mylar (0.08 mm) was positioned about 12 cm below the exit window, directly in the path of the scattered particles. This mirror directed emitted radiation onto a smaller spherical mirror at the side of the detector, and a photomultiplier tube (PMT) was placed on the opposite side at this mirror's focus. A wavelength-shifting coating was applied to the tube face to increase it's sensitivity to ultraviolet rays. Three drift chambers provided precision tracking. These chambers were used in the previous experiment NEII, but were found to be inefficient. Consequently, they were carefully cleaned and repaired before this experiment. Each drift chamber consisted of a 14 cm high box containing four wire planes separated by thin foils. Two wire planes were strung in the x direction and two in the y direction; the planes in a given direction were separated by only 09 cm and so were treated as a single plane. The arrangement of wires is depicted in Figure 25. The wire spacing was 1.0 cm, with cathode and anode wires alternating and charged to 500 V and 1850 V respectively. The cathode wires provided uniform field shaping, while the anode (sense) wires collected the signals induced by the 64 Anode (sense) wires Kapton foils ,. 'A I ' * 0 0 0 ? i ii*% * 010, 0 0 0 0 0! '11 --- * 0 00 0 a 0 0 I 0 0 0 Cathode (field shaping) wires Figure 25: 'The arrangement of wires in the 16 GeV/c drift chambers; each chamber consisted of two such pairs of wire planes, strung alternately in the x and y directions. ionization trails of fast charged particles. The chambers were equipped with gas input and output valves, and were connected in series to a flow system that circulated isobutane gas at a rate of 20 cc/minute. Each sense wire was attached to a channel of a common start TDC module; the 16 trigger (which does not include the wire chambers) provided the start, while each wire signal provided the stop for its own channel. Precise eent timing was provided by four planes of plastic scintillator bars, two segmented in the each of the x and y directions. The scintillators were 1.1 cm wide and their segmentation was used to assist with track identification. For this reason, their transverse edges were precisely mapped out relative to the wire chambers using tracks reconstructed. from the data. Each scintillator bar had phototubes attached at either end, and each of these was connected to both a TDC and an ADC. These measures permitted careful determination of the time a charged particle passed through each scintillator (see Section 35). At the top of the detector stack were two layers of 14 lead glass blocks each. They measured 10.6 cm along each horizontal side by 25 cm vertically, and were each viewed by a PMT. A.pion with momentum greater than I eV/c has energy loss similar to that of a minimum ionizing particle ( this is much less than that of a I 2 MeV/g/cm'); 65 GeV/c electron which loses energy predominantly by Bremsstrahlung radiation. In fact the measured electrons typically deposited all their energy in the calorimeter. The total signal deposited in the shower blocks during an event consequently provided a second means of particle identification. The segmentation of the shower blocks was also used in track identification, but the precision of the device in this context is sharply reduced by the transverse size of the electromagnetic shower that is precipitated by a high energy electron. This transverse size was found to reach a radius of about 4 cm at the top of the shower counter, and so energy was generally not localized to a single block in any row or column. 2.5 The 8 GeV/c Spectrometer GeV/c spectrometer is depicted in Figure 26. It is a QQDDQ spectrometer, which refers to the sequence of quadrupole and dipole magnets seen by a particle emanating from the target; it's optical axis rotates a total of 30' over the total magnetic path, culminating in a shielded detector hut inclined at this angle to the floor. The spectrometer is described in detail in ref. 35]. In the NE18 experiment, a new "large acceptance" magnet tune was adopted. This involved reversing the polarities of the first two quadrupoles from their normal acceptance" configuration, and resulted in an increase in solid angle of more 44 than a factor of 4 Like the 1.6 GeV/c, the GeV/c provided point-to-point focusing in the bend plane and line-to-point focusing in the non-dispersive (horizontal) direction. It should be noted that a different coordinate system is used to describe positions in the GeV/c than in the 16 GeV/c: z is directed along the magnetic axis, x points horizontally toward higher scattering angle, and y is defined to produce a left-handed coordinate system (and so points vertically upward at the spectrometer entrance). The GeV/c detector package is depicted in Figure 27 and consists of a Cerenkov 66 QC 0 0za Q WW ln 0a kA 0 z0 C D 11 &A J Ai O W -i 0.9 (n Figure 26: Side view of the 67 GeV/c spectrometer detector, planes of segmented scintillators, and 10 multi-wire proportional chambers (MWPC's). The 31 m long Cerenkov detector was filled with freon 114 (n=1.00140 I atm of pressure) only for the highest two Q2 settings. It was intended that the Cerenkov provide pion rejection at these energies, while at the lower energies the long drift distance of 386 m between the front and back scintillator planes would permit particle identification by measured particle velocity. In practice, the pion rate in the GeV was found to be sufficiently low relative to the accidental coincidence rate that pions could be efficiently analyzed as accidentals, and particle identification was not needed. Timing information was entirely provided by the five scintillator planes. The SF layer was segmented horizontally, into eight bars of 6 cm width framed by two of 15 cm width. The NBS hodoscope consisted of three closely spaced planes: the outer planes consisted of ten scintillator bars 48 cm wide by cm tall, stacked in two adjacent columns of five each, while the middle layer was formed from 22 cm wide bars placed side by side. The SM and SR planes were vertically segmented, into three 42 cm tall bars. Finally, the 10 MWPC, positioned between the SF and NBS scintillators, contained wires of I mm separation oriented in alternate planes at O' and 60' to the horizontal plane. The wire angle of alternate 60' planes also alternated to the left and right. The 00 planes were referred to as "P" planes and those at 60', as "T" planes; these labels refer to the planes' sensitivity to particle momentum and in-plane angle ("theta") respectively. A "magic gas" composed of 65.75% argon, 30.0% isobutane 025% Freon 13BI, and 30% methylal was circulated through the chambers. 68 r'Mf-^K I I I A /- -I n I A N e Chambers loscope entrance window of Cherenkov SM SR 384 cm I 0 I 1 M Figure 27: Diagram of the GeV/c detector stack. The detector positions are to scale, but their thicknesses are exaggerated. 69 2.6 Trigger Electronics 2.6.1 16 GeV/c Migger The formation of the 16 trigger is depicted in Figure 28. All 16 GeV/c detector elements except for the wire chambers participated in the trigger, after being grouped together in the following way. The scintillator tube signals were discriminated and Red by plane, producing signals XU, YU, XD, and YD. (The labels correspond to the plane names indicated in Figure 24.) These in turn produced SU (XU and YU in coincidence), SD (XD and YD in coincidence), and SC (a 34 coincidence between XU, YU, XD, YD). The shower counter singals were amplified by a factor of 10, and summed by FAN IN modules in stages to form PASUM and PBSUM (referring to the lower and upper shower planes respectively). These were summed in turn to produce SHSUM- Discriminators then converted PASUM to two signals: PA Lo and PA Hi, with threshold settings of 155 mV and 500 mV respectively. SHSUM was discriminated only once (at 600 mV to produce SH. The Cerenkov output was immediately discriminated to produce signal CK. All discriminated signals had widths of 20 ± ns. Two electron triggers were defined, both intended to fire at every electron while providing some degree of pion rejection. EL Hi was formed from a 33 coincidence of PA Hi, SC, and SH. This trigger provided the stronger level of pion discrimination, using both the summed information from all shower blocks (SH) and a relatively high requirement on the energy deposited in the first layer (PA Hi). The latter condition was motivated by the fact that the rate of energy loss of hadrons in matter peaks toward the end of the hadron's path (described by a "Bragg curve" 36]). Electrons radiating away their energy, in contrast, do so with -dEldx - E and so should deposit relatively more energy in the first shower layer. These shower counter tests, however, are not 100% 70 efficient for electrons, because of cracks between the shower blocks, the low energy tail of electron energy loss, and a sporadic malfunction in one of the shower block ADCs (PA4). Thus a second electron trigger, EL Lo, was also used. EL Lo required a 23 coincidence between SU, SD, CK, and PA Lo. Thus, the shower requirement is weakened, and was not required for the formation of the trigger. The high efficiency of the Cerenkov and scintillator planes in detecting electrons makes this test 100% efficient to well within experimental statistics. The final electron trigger EL20 was formed from an OR of these two signals and given a pulse width of 20 ns. The 16 GeV/c pretrigger was formed by an OR of three signals - EL20, RANDOM, and PION PRESCALE -, arriving in coincidence with the beam gate. RANDOM was generated by a low rate pulser, also vetoed by the inverted beam gate. Its purpose was to provide "empty" events where hardware information such as ADC pedestals could be periodically monitored. PION, designed to provide a sample of pion events, consisted simply of SC vetoed by the Cerenkov CK. The high PION event rate required it to be prescaled before being included in the pretrigger. This was accomplished by forming PION PRESCALE from PION in coincidence with the prescaled beam gate (by up to a factor of 2'). Finally, PRETRIGGER 16 was converted to TRIGGER 16 by forcing the final trigger rate to at most one event per spill. This requirement is imposed by the processing speed of the data acquisition. It was accomplished by vetoing the pretrigger with a long (1.0Ms)gate started by a delayed copy of the pretrigger. This gate effectively blanked out any further pretrigger signals arriving before the next spill. 2.6.2 8 GeV/c and Coincidence rh-iggers The formation of the GeV/c and coincidence triggers is shown in Figure 29. Of the 8 GeV/c detectors, only the scintillators participated. The PROTON signal was formed 71 Figure 28: Formation of the 16 GeV/c trigger. from a coincidence of the 3 planes SF, SM, and SR. Another trigger, called 2/3 or PION, required only 23 of these planes; its purpose was to provide an event sample with which to test the efficiency of individual scintillator planes. Since virtually any particle entering the GeV/c hut would cause any scintillators it encountered to fire, the 2/3" event rate was extremely high and had to be prescaled to prevent prohibitive dead times. Factors of 2 2 to 2' were used, for all but the 11 targets. The PROTON signal did not contribute directly to the final trigger. Instead, it was used to form the coincidence triggers. The 16 pretrigger started 2 gates, one 100 ns and the other 31 ps long. A PROTON trigger arriving within the short or long gate cause a COIN or LONGCOIN signal respectively. COIN provided the true coincidence trigger; LONGCOIN was defined to provide a selection of PROTON singles. The PROTON signal was delayed by an amount that was adjusted at each kinematic setting to position the coincidence peak roughly 20 ns after the gate start. Note that the leading edge time 72 TYAWER memoC.& Figure 29: Formation of the GeV/c and coincidence triggers. of COIN and LONGCOIN was that of the PROTON signal. The GeV/c pretrigger was then formed from and OR of COIN, LONGCOIN, 2/3", and RANDOM (shared with the 16). As in PRETRIGGER 16, PRETRIGGER 8 was vetoed by the inverted beam gate, and TRIGGER (7 was formed by vetoing PRETRIGGER on a delayed widened s) copy of itself. Two coincidence TDCs were setup to measure the relative time between the electron and proton triggers. TDC8 was started by the GeV/c COIN trigger and stopped by the 16 GeV/c EL20 trigger, while TDC16 was started by EL20 and stopped by COIN. 2.7 Data Acquisition Signals from all detector elements were delivered to the counting house on heliac cables and processed by a DEC PDP-11 computer. All 16 and GeV/c triggers were recorded to tape. At the beginning and end of each run comprehensive begin and end headers 73 were recorded, and every - 3 minutes, a checkpoint was stored containing a summary of charge and scaler information. The online analysis package available at SLAC ran continuously on a VAX cluster during data taking and provided real time information on not only the detector systems but also the results from single-arm tracking. The data acquisition package was also modified to write coincidence events (COIN triggers) to compressed disk files. The relatively small size of these files permitted coincidence information from the entire experiment to be stored and made available for analysis by the oine coincidence analysis program. This arrangement permitted us to monitor the coincidence rate even at the highest Q kinematic setting, when the true to accidental rate was so low that no timing peak could be discerned from the raw coincidence TDC distributions. (see Figure 210). 2.8 Kinematics The kinematics of the experiment were chosen based on several considerations. A broad range in Q2 over which to study quasielastic (ee'p) scattering was desired, and in par- ticular, we wished to take data at the maximum possible value of Q in order to search for colour transparency. The limiting factor in this case was the GeV/c momentum, which could only be set to 45 GeV/c. Beyond this, the solid angle acceptance in the large acceptance tune dropped away sharply, as of Q2 Consequently the maximal value that the experiment could achievewas 68 (GeV/c)'. We also wished to sample as much of the nuclear spectral function as possible, and in particular to detect p up to the Fermi momentum for each nucleus and at each Q. This was achieved by varying the proton angle at each to see the peak of the spectral function at p Q2 around the value required = . In so doing, we employed so-called perpendicular kinematics: changing the angle of the proton changes the component of 74 r I"II 1 .... I.................................... . . . . . . . . . I Raw TDC, spectrum "I V) -+l C I 0 0 -1-- I'l I I HI I 0 . ..... . ..... . .... -I5 0 - b- IT . . . . . . -F- . . . . . . . I. . . . . . . . . I0 t8 - t 1.8 (ns) to - ,, (ns) 5 I 10 20 15 V) -.1 C I 0 0 ic C Figure 210: Raw coincidence TDC and corrected coincidence timing spectra for 197 Au at Q2 of 68 (GeV/C)2. 75 the pm vector. perpendicular to q in the scattering plane, while the detection range of the other components remains unchanged. This choice of kinematics is the motivation behind the sign we apply to the magnitude of the pm vector (see Section 1. 1). As Q (and so P') increases, a kinematic focusing of the "Fermi cone" of proton momenta occurs. This refers to the fact that a given change in p, requires an ever smaller change in P' as p' increases. Thus a given setting of the proton spectrometer momentum provides a larger detection range in p, at higher Q, and the number of proton angle settings required to cover the Fermi cone decreases. The relatively high energies of the experiment and large momentum bites of the spectrometers ensured that missing energies up to the pion mass at 140 MeV would be detected. Figure 211 depicts our experimental phase space in E, and p, for carbon at Q = I GeV/c)', and graphically illustrates the effect of Op,on the missing momentum coverage. Our kinematic settings are summarized in Table 23. 76 Table 23: Kinematic settings used in NE18. Q2 Targ ts (GeV/c)' 1.04 lH c f Et pI 0, OP (GeV) (GeV) (GeV/c) (deg) (deg) 2.015 1.41 1.39 1.24 1.20 37.3 35.5 1.36 1.28 38.8 3.188 1.47 2.49 2.45 49.0 47.7 4.212 1.47 3.56 3.54 54.2 53.4 5.120 1.47 4.49 57.0 56.6 43.3 35.0, 37.8, 40.6, 43.4, 46.2) 49.0, 51.8, 54.6 43.4, 46.27 49.0, 51.8, 54.6 35.9, 39.1, 41.3, 43.5, 46.7, 50.2 26.5 27.7 27.7, 30.5, 33.3 27.7, 30.5 19.6 19.5 18.5, 20.9, 22.6 20.9, 22.6 20.9 15.9 15.9 15.9, 17.3 15.9, 16.7, 17.3 16.7 Fe, Au 1.21 2H 3.06 lH 2H 5.00 C, Fe Au lH 2H 6.77 c Fe Au lH 2H c Fe Au 77 400 - ' ' :....... . . ,=55' . . . . .3... 300 . - - -=52' 0,,=49' 0,=46' 0,=43 2 a 200 100 U '-I 0 I ....... .... Q) ' ::. . . . . . . :: . , ................ .. ::: .. . i E a_ I I I: : . ... . : : . -I I I I I I -100 :' :......... gig . :: : -200 - : : -300 -1( I . . . . . . . .. : ... : I ' ' 1 )O -50 1 50 0 I 100 - I-- 150 I 200 E M (M eV) Figure 211: The NE18 experimental phase space in E and p for the carbon Q = I (GeV /C)2 measurement. The eight diagonal bands that can be distinguished in the figure correspond to the coverage of each of the proton spectrometer settings. 78 Chapter 3 Data Analysis 3.1 Overview The reduction of the raw experimental data to useful physics information took place in two stages. First, detector information was read in event by event from the coincidence disk files described in Section 27, and the raw measurements from each spectrometer hut were reduced to tracks. Using optical models of the spectrometers, these were reconstructed to particle vectors at the target. The tracking algorithms were modified versions of programs developed at SLAC over the lifetime of the detectors. Secondly, the nuclear structure quantities E, and p, were determined from the single particle vectors, and true coincidence events were identified. This step included corrections to the single particle vectors for their energy loss in the target, fitting of the coincidence timing peak, and random subtraction. 3.2 Tracking The tracking algorithms in both spectrometers basically proceeded by constructing tracks from all combinations of wire chamber hits, then eliminating candidates using a sequence 79 of cuts and purges. At the end of the sequence of track reduction tests, it is possible that several equivalent candidates remain; these are referred to as multiples. The tracking was operated in the so-called "one track" mode, where a single track was always identified from a collection of multiples by selecting the track which arrived first in time. The possibility of the wrong track being chosen from a set of multiples and the motivation for this selection criterion are dealt with in Section 32.3. A "no track" correction is applied to the results to account for events where an actual coincidence event occurred but no candidate track was found. This correction is determined by examining the raw coincidence TDC spectrum for events where a given spectrometer was unable to find a track, and calculating the coincidence content of this TDC peak. The corrections used are < 3 3.2.1 and are described below. 1.6 GeV/c ']Tracking At the outset, a cut on the Cerenkov ADC value was imposed on all events considered bythe 16 tracking algorithm. This pion rejection measure was made possible by the excellent efficiency of the Cerenkov (>99.9%). Tracks were first constructed by considering all possible combinations of wire hits, in the x and y directions separately. Particles passing through a wire plane generally caused a pair of wire hits - one on either side of the track's position. Occasionally, however, only a single wire hit would be recorded. Consequently two different criteria were required for the construction of a candidate track from the wire hits: a "good" track was constructed from at least two pairs and a total of or more wires, but any track consisting of at least 4 wires was also considered. If a "good" track was found for a given event, only "good" tracks were considered for that event. Candidates were also required to point inside the vacuum chamber exit of the spectrometer and within the boundaries of the shower counter. Furthermore, at each 80 scintillator plane in which both phototubes of any scintillator bar fired, tracks were required to pass through a scintillator bar in which edher tube fired. At the end of this procedure, the number of possible x and y tracks were counted, and if either was greater than one, a prging algorithm was called to decide between them. A series purges was applied to the tracks in the sequence described below, until only a single candidate remained. At each stage, the x and y tracks were combined to form candidate pairs, and it was these pairs (rather than individual xly tracks) that were discarded. Any purge that discarded all tracks was dismissed, so that no events with candidate tracks were discarded at this stage. (This is with the exception of the test on reconstructed quantities, which was applied as a hard cut). shower counter The software gains applied to the shower counter ADCs were tuned so that a signal of "I" in any block corresponded to the electron energy loss peak for that block. Since most of the electron loss occurs in the lower ("A") layer, the gains o the A blocks were lower than those in the block. This test required that the sum of energies deposited in shower blocks along a track be greater than 0.6. This threshold value was determined by applying the test as a hard cut and progressively lowering the value until no significant gains in the final number of coincidences was obtained. The determination of which blocks could be considered to lie on a track involved the use of a slop factor of 4 cm, to account for shower spreading; this value was determined in a similar way. It should be noted that a few of the shower counter ADCs malfunctioned sporadically during the earlier part of the experiment. For these events, the shower purge was not used. This choice was supported by examining runs where the malfunction affected a limited percentage of events, and verifying that the number of extracted coincidences was the same if one neglected this test or applied it as a hard cut and later corrected 81 for the fraction of bad ADC events. target cuts Candidate tracks were reconstructed to target at this stage (see Sec 33) and loose cuts were placed on the reconstructed vectors to eliminate very improbable tracks. These cuts were about a factor of two larger than the spectrometer acceptance in each coordinate, and so did not bias the event sample in favour of tracks away from the acceptance edges. purging of "nearby" tracks Due to the high event and background rate in the chambers, a handful of wires which were either noisy or dead (and therefore forced on in software), and the relatively spare requirements for the construction of a track (e.g. only 2 chambers required), a common source of multiple tracks was the construction of two or more "nearby" tracks which actually corresponded to the same event. To eliminate spurious choices, tracks within 2 cm of each other at the focal plane were defined as "nearby" and those with less the highest number of wires and pairs in the group were discarded. Finally, a purge designed to identify the most common patterns of spurious "nearby" tracks was employed. These patterns included tracks separated by no more than a wire at one plane, tracks constructed using hits in only 2 out of 3 available chambers, and tracks involving a dead or hot wire. The graphics capabilities of the oine software, which enabled single events and associated tracks to be displayed on screen, were extremely helpful in the design of these tests. The "no track" correction factor for the 16 GeV/c was found to vary from I to 2 and is listed in Table 3.1. 82 3.2.2 8 GeV/c Tracking The tracking algorithm used for the eV/c was very similar to that used for the 1.6. However, it was called repeatedly, with three increasingly lenient track criteria, until at least one track was found. First, a track was required to include at least 3 P chambers, 3 T chambers, and 7 chambers in total. Second, 2 P 2 T, and 6 total chambers, and finally only 2 P 2 T, and total chambers were demanded. Candidates were determined first in the y (vertical) direction, using only the P chambers, then companion x tracks were identified using the y tracks already found to resolve the horizontal positions indicated by the slanted. T chambers. The purging tests in the GeV/c were applied in the following sequence. We should note that a hard cut on the extracted of the track was applied at the outset, as a pion rejection measure. X2 cut A maximal X2value of 30 was required. fiducial cut Cuts were placed on track position at the SF, NBS, and SR scintillator planes which defined a fiducial (contributing) region in the hut. scintillators on track Tracks not matching the information from planes which fired at least one tube were purged. Only the SF, SM, and SR planes were used in this test. NBS hodoscope Tracks not matching the NBS hodoscope information in at least one direction were purged. If no tubes fired in a given direction, the track was passed. The relative looseness of this test was due to the inefficiency of the NBS scintillators. scintillators in combination Tracks that failed to satisfy either a strict scintillator or a strict NBS cut were purged. These strict tests were each about 1-5% inefficient, but the efficiency of the OR is about 0.1%. 83 x-dth correlation Due to the line-to-point focusing of the GeV/c, the horizontal po- sition and angle at the focal plane of a track emanating from the target should be closely correlated. Tracks falling outside this correlation band were rejected. target quantities As in the 16 GeV/c purging, loose cuts on target quantities (well outside the spectrometer acceptance) were applied. The reconstructed position at target along the beamline is more accurate in the GeV/c than in the 1.6 and so a loose cut on this position based on the length of the target was also applied. For the GeV/c, the vast majority of "no track" events registered as coincidences by the coincidence TDC were caused by the fact that the wire chambers do not quite cover the acceptance of the GeV/c hut, while the trigger-forming scintillator planes did. This effect is taken into account by the PWIA Monte Carlo which is used to normalize all of our results. We apply a % systematic uncertainty to account for any tracking losses. 3.2.3 Multiple h-acks After the sequence of purges described above is applied to a set of tracks, there may still remain several equivalent choices; in this case the event is referred to as a "multiple". In the "one-track" mode, a single track is always selected from such a sample. This is accomplished by determining each track's time relative to the 1.6 or GeV/c trigger, after projecting back to the focal plane. The track with the earliest time is then selected. The reason for this choice is that tracks in a multiple sample very often share one or more scintillators; since the scintillators are used to determine the event time and since only the first track to pass through a scintillator will have the correct time assigned to it, selecting the track that is earliest in time avoids the possibility of losing coincidences because their coincidence time has been computed inaccurately. The number of multiples in the GeV/c is in general low 84 less than 2 for all kinematics except the deuterium runs at Q = I GeV/c)' and at forward proton angles. At these kinematic settings, the number of background events in the GeV/c increases dramatically, and up to 10% of the analyzed events are multiples (at the most forward angle, 08= 36"). This high rate of multiples can be explained by the rate recorded by the PROTON scalers; in other words, the multiple tracks correspond to real trigger-causing particles. The! loss of coincidence protons due to earlier background protons arriving in the same beam burst can then be calculated as a component of the computer deadtime. This is examined in detail in Section 36.2. The fraction of multiple events in the 16 GeV/c, however, is much higher - ap- proximately 5% at low Q settings, and reaching 10% during some Q = 7 GeV/c)' runs. The 16 GeV/c is always subject to a high pion background rate, and at high Q, this rate becomes several hundred times larger than the electron rate; the source of the multiples is clearly pions accompanying trigger-causing particles into the spectrometer. The question is, do the multiple events contain the same fraction of coincidences as the single track events, or is there something innately "bad" about them? The answer lies in the purge on the amount of shower energy found along each track. The large majority of the multiple events are caused when all the candidate tracks failed the shower purge; in this case, the purge is discarded and the tracking must use only its remaining tests to refine the sample down to a single choice. Although a Cerenkov cut is placed on all events analyzed by the 16, there is a substantial possibility that a pion produced a knock-on electron which then caused the Cerenkov to fire. To determine a correction for the 16 multiples, the fraction of coincidences found in single-track samples which passed and failed the shower purge was computed. This fraction is indeed much larger for the sample of good shower events. The multiple tracks were then discarded, and those which passed and failed the shower track test were counted separately. Different corrections 85 were applied for each of these sets of multiples, using the fraction of coincidences found in the single-track samples. In practice, the "one-track" method of selecting the first event turns out to be very successful at choosing true electrons. The net corrections applied to the "one-track" results to compensate for multiple track losses is at most 2. Upon reflection, this is not terribly surprising. The rate of multiples in the 16 is much larger than the number one would expect from the rate of 16 EL20 triggers, and so it is clear that the extra tracks are not pions which leaked through the electron trigger. They are therefore not the particles which caused the trigger. Since the 1.6 trigger starts all the TDC's, the pions will in general have later times at focal plane than the true electron. 3.3 Reconstruction of Events to Target Tracks determined in each spectrometer were reconstructed to particle vectors at the interaction point by means of matrix models of the spectrometer optics. During data taking, the online analysis used matrices determined by TRANSPORT 37] a program developed at SLAC which computes optical matrices describing the passage of charged particles through a magnetic system. Input to the program is provided by "decks", files describing the position, shape, and properties of all magnetic elements in the system. During the post-experimental analysis, however, a powerful tool for understanding the spectrometer optics was available in the form of elastic data. The matrix elements were fit to this data using an iterative procedure, as described below 38]. The technique involves correlating target quantities which are known (by virtue of fixed target positions, collimators, or kinematic correlations) with focal plane quantities (or combinations of them). Any correlation indicates the need for a matrix element; the element's value is determined by eye from the slope of the correlation. The necessity for doing this by eye rather than with a fitting procedure comes largely from the presence of radiative tails, 86 which distort the shape of an elastic stripe, for example. By hand, the true stripe can be identified. Two coordinate systems are used in this discussion. The first refers to particle coordinates at target, with Z pointing downstream along the beam line, Y directed vertically upwards, and X defined to form a right-handed system. Z of Z onto the spectrometer entrance plane. The angle refers to the projection is the difference between the in-plane angle and the central spectrometer angle, with positive angles. at larger scattering is the out-of-plane angle, with positive (Dcorresponding to angles above the plane of the lab. Positions at the focal planes (in the hut) are described instead using the TRANSPORT convention: x along the dispersive direction toward larger momentum, z along the magnetic axis in the direction of particle travel, and y forming a right-handed system. Track angles in the x and y directions (relative to z) are denoted dx and dy respectively. 1. The Z matrix elements are fit first, using the 15 cm dummy target cell. This target consists of two aluminum plates at precisely known positions and with effectively zero width. The Z ly, Z Idy, and ZLJO (offset) matrix elements are fit so that the reconstructed positions are correct and the resolution is optimized. In practice, resolution in this coordinate was poor (FWHM > cm), and only very loose cuts on target position were used in the analysis. 2. (D16 matrix elements are determined using the 4D16distribution for 'H. The center must be at the edges must correspond to the vertical edges of the fixed slit at the front of the 16 GeV/c, and the distribution should be filled uniformly because of the idependence of the cross-section on this coordinate. Also used here were checkout runs where the movable slit at the spectrometer entrance was positioned at ± 21 and 27 mr. 87 I % and DN matrix elements are tuned using H and requiring that the out-of-plane component of pm is independent of these target angles. 4. For 'H data, E = - - E - M, - T.. is fixed at the binding energy of 22 MeV. Selecting a sample of events with p less than about 10 MeV, the effect of TI,, on E, can be kept to less than about MeV so that, to this accuracy, E, depends only on the difference of spectrometer energies. Such an event sample can be used to tune the secondary primary dispersive 68 and 16 matrix elements (i.e. other than the x elements), because any correlation between E, and dx, for example, must be due to a 1dx matrix element. 5. The 'rimary dispersive blx) and matrix elements were fit using a series of H runs, performed during checkout, where the central momenta of the 16 and 8 GeV/c were varied independently in steps of 3. eV/c Handles on these matrix elements are obtained using the exact correlations between all scattering angles and momenta determined by elastic kinematics. A momentum scan moves the elastic peak across the spectrometer'sfocal plane in known steps and thus provides information on the spectrometer dispersion. 6. Matrix elements were checked using an angle scan of the GeV/c, also done on H during checkout. 7. Offsets in all target quantities can be determined using H data, available at all Q2. These offsets were fit at all Q2 , and were found to vary somewhat. Varying offsets in the beam energy as its value changes, and offsets in central spectrometer momentum as the magnets are retuned are believed to be responsible. Ultimately, only first and second-order matrix elements were required in the 16 GeV/c, but three third-order elements had to be introduced in the 88 GeV/c. These were Olyx', Oxy', and a very small term in bxy'. If one uses the width of the extracted E, and p. peaks in hydrogen as a measure of the net resolution of the reconstruction, one finds that the spread in beam energy (which contributes to each of these measured quantities) dominates the resolution at all Q. This is determined using the fact that elastic kinematics are fixed by the beam energy and only one other scattering angle or momentum. Using this overdetermination, one can compute the beam energy for each event using, for example, the proton scattering angle and then examine the widths of the E,.,, and p, spectra. The resulting resolution is substantially less than that incurred by 'dEbe.,.1Ebe.,.. 3.4 The reconstruction optics can thus be considered well-known. Determination of E and m Before E, and p were determined for each event, two corrections were applied to the measured particle energies to obtain a good estimate of their values at the scattering vertex. The ionization losses of the particles as they passed through the materials listed in Table 22 were calculated using formulae based on the Bethe-Bloch equation. For particles heavier than an electron, this equation gives the most probable rate of energy loss as dE = 03071z2 dt In 1 A 2m'02 2 I( - - _ 02 (3.1) 2) where dEldt is in MeV g CM-2 , z is the charge of the particle, m, is the mass of the electron, and I is an ionization constant. I depends on the material being traversed: for Z > I = 16Z'-' eV, while for liquid correction for the charge screening provided by atomic electrons was neglected because H2 and D2 I = 21.8 eV. The density it contributed less than 5% at our highest proton energy. For electrons, the Bethe-Bloch 89 formula must be modified to take into account the substantial deflections the electrons undergo because of their lighter mass, and the fact that their interactions are now between identical particles (incident and atomic electrons). This calculation leads to the following expression in the limit 1: 391 dE = 03071 Z 1 19.26 dt A2 I N ] kpjj (3.2) Here, t is the material thickness in g cm- 2 and p is its density in g cm- 3 The incoming and outgoing electrons are affected not only by the struck proton, but are also accelerated and decelerated respectively by the Coulomb field of the spectator protons in the target nucleus. For an electron in either arm experiencing a hard scattering at a radial position r inside the nucleus, the total change in potential energy it sees between this position and a location infinitely far away is given by AEc (Z _ )e2 41rEORO 3 2 2 (3.3) 2 where R is the charge radius of the nucleus, given approximately by Ro = 1.18A1/3 fM. Averaging this over a sphere of radius Ro, one finds a mean energy change of 6 (Z - )e2 2K-E = 5 47reoRo- 3.5 (3.4) Timing Thin plastic scintillator bars produce signals characterized by quick rise times that make them ideal timing devices. All of the scintillator times found along a given track were used to determine the time of event production at the target, after applying careful corrections. These corrected single-arm times were then combined with the information 90 from the coincidence TDCs to provide a distribution of relative event times. In each spectrometer, the time at which each particle track crossed the focal plane as well as the particle's speed were determined by fitting these quantities to the particle's trajectory and. recorded times at the scintillator planes. The timing at each plane was corrected for wo effects. First, the scintillation light produced when a particle passes through the scintillator takes time to travel to the phototubes. The difference between the times measured at either end of a scintillator can be used to correct for this, using the light travel speed and the length of the scintillator. The light travel speed is about 4 cm/ns, and was determined for each scintillator bar by a fit to the "sum time" distributions. (The sum time refers to the sum of the times recorded at either end of a scintillator). Second, one notes that a phototube signal with a larger pulse height will cross its TDC discriminator threshold sooner than one with a smaller pulse. Accordingly each TDC time was corrected by a function of the ADC signal, with parameters fitted to optimize the timing resolution. Once the time at focal plane is determined, orbit corrections for the exact particle path through the spectrometer must be performed to determine the event time at target. This path length was calculated from a track's coordinates at the target using a series of matrices base n TRANSPORT models of the spectrometers. The matrix elements were fit to the data to optimize the coincidence time resolution. The coincidence (relative) time is given by the simple formula ctime = tproton and -TDC8 - tle where tproton and 6 GeV/'c triggers, after projection back to the target. The delay time tlex (3-5) + tdelay, refer to the proton and electron event times relative to the tdlay was fit at each kinematics to center the coincidence timing peak. We should note that two 91 k I k 10 R Window B Window ctime C Window 31: Diagram representing coincidence timing peak and windows used in selection of "true" events. Figure coincidence TDC's were available: "TDC8" started by the proton trigger, and "TDC16 started by the electron. TDC16 malfunctioned during the latter half of the experiment, and so TDCs is used throughout the analysis. However, a non-linearity in TDC8 at small values (which was corrected for in software) caused the coincidence peak to be set too near the TDC underflow. Using the results from TDC16 at the first two Q settings, a correction factor of 1.5 for the portion of the coincidence peak lost by TDC8 was determined. This correction is referred to as deff in Table 31. The resolution of the experiment is demonstrated in Figure 210: the coincidence peak had FWHM widths of around 06 ns. Figure 31 illustrates the procedure used for random subtraction. For each kinematic setting, a range over which the ctime background was flat (the "R window", typically ± 8 ns) was determined, and the ctime peak was fit to a Gaussian with a flat background. 92 The extracted FWHM produced two further timing windows, the "C window" at ± 2 FWHM, and the "B window" at ± 1.5 FWHM. Histograms of the event distribution in E, and p, were taken from the window, to increase the coincidence to accidental rate, but were corrected for the randoms and for the total number of events in the wider C window: Ntrue(EmIpm = [NB(Em7pm - R,.t,,ATBNR(E,.7pm)]C7 C Nc(Em7pm - RrateATC (3-6) NB (E., pm - RrateATB' where R(E,7pm) denotes the limits of the arrays, AT refers to the size of the timing windows, and the random rate is given by Rrate ER(E,,,,p NR(Em, pm) (3.7) ATR 3.6 Corrections The correction factors applied to the data are summarized in Table 31. The NOTRK, multiple track, and coincidence timing efficiency (deff) corrections have already been discussed. The computer deadtime and proton absorption corrections are described below. The electronic deadtime corrections discussed in Section 36.1 are not applied directly to the data but instead are needed to correct scaler rates used elsewhere in the computer deadtime calculations. 3.6.1 Electronic Deadtime Corrections A scaler counting logic pulses of duration -r and mean rate r will record counts at a smaller rate T, The reason for this is that two counts arriving within a time interval less 93 than will overlap and only one will fire the scaler. A scaler is a so-called paralyzable device in that it does not "reset" itself after each count, and successive overlapping pulses consequently cause deadtime to accumulate. (This is unlike a non-paralyzable device, such as a simple data acquisition system, which is dead to incoming events while it is processing, but is ready to accept new information as soon as it is finished with the current event.) The behaviour of a scaler can be analyzed in terms of Poisson statistics. The probability of finding n counts of mean rate r within a time interval t is P (n = (rt )n,-rt (3.8) n The probability of finding no events in an interval t is then from which we can infer the (normalized) probability distribution for the time intervals between events: P(t = re -rI (3-9) A paralyzable scaler will only record events that occur more than 'r apart. The fraction of counts recorded by the scaler is then 00 r re tdt e -rT (3-10) What one actually measures however are not rates but numbers of counts integrated over a run. To convert between a rate r and a scaler measurement n, one needs the effective total beam time T delivered during the run (n =_rT). Basically, T = TpjnNrpjjj, the product of the number and length of the beam bursts, but one cannot simply use Tpjl = 1.6ps since the spills certainly did not possess a perfectly uniform charge distribution. One can effectively avoid this problem by sending a signal of interest to two gate 94 --- generators of ifferent widths and then to separate scalers. Specifically, if two rates , and F2, are measured, the true rate can be extracted as follows: ii = _nr ne ji2 n- - T . T (3-11) n2-r In other words, the electronic deadtime correction for fi is elec KIr n2,r (3.12) Several scaler rates needed in the analysis were indeed measured with more than one pulse width, but many were not. To deadtime correct these, one requires the total effective beam time T so that Equation 310 can be applied. Using any pair of scaler rates and equations 39 and 311 one can determine this value: ii2 n2, In (3.13) Equation 311. then yields the true number of events, but cannot be solved analytically. Due to the high rates of some of the event types (e.g. up to in the 23" triggers per spill GeV/0, the equation cannot be approximated by a first order expansion inrr. However a simple iterative procedure can be used to determine n. 3.6.2 Computer Deadtime Corrections Computer deadtime refers to the loss of events incurred by the fact that the data acquisition system can process at most one event per beam burst. The computer deadtime correction is defined to be the ratio of the actual number of events of interest to the number analyzed. Generally, this is given by the ratio of pretriggers to triggers, but because of the design of our coincidence trigger this is not an appropriate approach here. 95 The experimental trigger was formed from the OR of the 16 and GeV/c triggers; the 1.6 trigger consisted of EL20, PION PRESCALE, and RANDOM, while the trigger was composed of COIN, LONGCOIN 23 (PION) PRESCALE, and RANDOM. The random trigger is not included in the deadtime calculations because its very low rate < .0.002/spill) gives it a negligible chance of blocking a coincidence event. LONGCOIN can also be neglected because it cannot occur without COIN and so cannot block coincidences either. Three sources of lost coincidences remain: the 16 trigger rate, the prescaled 23 rate in the GeV/c, and the chance that a background proton blocked a coincidence proton in the formation of the trigger. 1.6 Yigger Rate Every coincidence requires an electron trigger. Thus if we neglect any complications from the GeV/c side, our deadtime correction is simply the total number of EL20 events over the total number which formed the trigger. Note that the possibility of PION events blocking electrons is taken into account in this ratio, by the denominator. The number of electrons forming triggers is directly measured by a scaler, E40-V, which counts EL20 events (widened to 40 ns pulse width) vetoed by an inverted 16 trigger (so that the only events counted are those which coincided with a 16 trigger). The total number of electrons can be determined via Equation 311 from the scalers EL40 and EL80, measuring EL20 events widened to pulse widths of 40 and 80 ns respectively. From now on, a deadtime corrected scaler measurement will be indicated with an asterisk, as in EL40*. A better estimate of the number of real electrons (without pion contamination) is provided by the CAB20-80 series of scalers, which measured coincidences between CK (Cerenkov), PA Low, and PB. The Cerenkov requirement, in particular, is the same as that present in the 16 analysis. However, the ratio 96 CAB40* CAB40-V ought to be equal to EL40* EL40-V or indeed to a similar ratio for any component of the total 16 rate. This is indeed the case, to within 1%. To summarize, the 16 trigger contribution to the coincidence computer deadtime is Kc Prescaled 23 Miggers in omp,1.6 EL40* EL40-V CAB40* CAB40-V' (3-14) GeV/c The only remaining component of the experimental trigger able to block COIN's is the 2/3 trigger of the GeV/c (also referred to as PION). Designed to provide a check on scintillator efficiency, this event type requires only a coincidence between two scintillator planes and can count at a tremendous rate (up to 8/spill). It was consequently prescaled, by blocking te 23 from the GeV/c pretrigger during a fraction of the beam bursts. During the reaining 1-,E)bursts, coincidence triggers might be blocked by earlier 2/3's. Let be the computer deadtime correction for these bursts. The total livetime unaffected by the 23 rate is then + yielding a net, contribution to the total deadtime of Kcomp,2/3= (I _E + To evaluate tc, we consider only the fraction (3.15) of bursts where 23 might contribute to the pretrigger. Further, let rOIN and r2/3 be the true COIN and 23 rates respectively. In the absence of 23 interference, we would expect the number of involving COIN to be ,ENpjjj(l - -rC01NT6i11). 97 eV/c triggers The actual number of COIN triggers, however, is PIPRTRIG rCOIN r2/3 where PIPRTRIG is a scaler measuring the number of triggers involving 23. This is understood by observing that every COIN requires a 23; the number of COIN signals participating in the trigger is then just the number of 23 triggers times the fraction of the 23 rate provided by rcOIN- PIPRTRIG itself is given in Poisson statistics by PIPRTRIG and so our correction (3-16) N.Pill( - e-12/3Tspill) can be written r2/3 rcolN The rates r2/3 and rOIN -rCOINTspill) (1 (3.17) e-12/3T-pill can be determined from the scalers PION and COIN respec- tively, after deadfirne correction and division by the total effective beam time T derived earlier. The result is then COIN' PION* Tr2/3 COIN* TrcOIN PION* (I K COIN* I e N.pill - PION* (3-18) Npill As a consistency check on the Poisson relations employed and on the scaler measurements themselves, the COIN sand PION values can be replaced by equivalent expressions computed from other scalers. Note that the latter concern - the accuracy of the scaler readings themselves - is largely due to the possibility of double-pulsing in scalers counting pulses of 20 ns width. The raw signals produced by photornultiplier tubes and the like generally consist of an initial voltage spike followed by a long tail. Secondary voltage 98 spikes may occur along this tail, and superimposed on the partially decayed tail 20 ns or so after the nital spike, may be of sufficient height to cause a discriminator module to fire twice. Most logic pulses were in fact about 20 ns long. Some scalers however, such as the CAB40-CAB80 series, were immune to this problem because the decay of a voltage tail after > 40 ns is sufficient to prevent superimposed spikes from reaching discriminator thresholds. Our approach is to use the redundancy of the GeV scalers to provide alternate measurements for necessary rates; the difference between the various results is then included in the systematic uncertainty of the experiment. From Equation 316 and 318, PION* = -Npili In I PIPRTRIG (3-19) ,ENrpin Note that the use of PIPRTRIG, which counts at most once per beam burst, avoids the need for an electronic deadtime correction. The coincidence rate, similarly, can be estimated fror CIN-V, which measures the number of COIN causing GeV triggers and also counts at most once per spill. For bursts where the 23 signals are blocked, the probability of a COIN trigger is I -rCOINTpill _ while for the rearnining bursts, the total number of COIN triggers is PIPRTRIG rcolN PIPRTRIG PION* r2/3 rcOINT. Consequently, COINV = Npju(l _ _ -rcOINT.Pill) +PIPRTRIG rcOINT. PION* 99 (3.20) To extract rcOIN readily from this expression, one can take a first order approximation of the exponential, assuming rOINTpi < 1. In practice, this product is at most 0.1, making the approximation valid to better than 1%. The result is an alternate expression for COIN*: COINV COIN* (3.21) PIPRTRIG' PION- Using these various expressions as input to Equation 315 and 318, one seeks to compute an uncertainty on K ,,p,2/'. The correction itself is only significant A.5%) for hydrogen and deuterium runs at Q2 < generally set between I and 2'. making the contribution of (GeV /C)2, where the prescale fraction. was For the remaining runs, was set between 2 8 and 2`0, to K"'P,2/' extremely small. For hydrogen and deuterium, the uncertainties were at most 1%, and this only at those few kinematic settings (forward proton angles) where K"'P,2/3 reached 1.10. (High 23 rates are due entirely to background processes, which vary inversely with scattering angle and momentum transfer). Blocking of Coincidence Protons A COIN trigger was formed when a PROTON trigger fell within a 00 ns gate opened by the EL20 trigger. The proton was delayed so that the coincidence peak fell - 20 ns after the opening of the gate. Note that a COIN trigger was also produced if the delayed proton signal was already present when the electron arrived. Consequently the 20 ns width of PROTON must be included to determine the total time window before the peak in which a proton could fall and produce a coincidence. This arrangement is illustrated in Figure 32, which shows a sample spectrum recorded by TDC8. KcompPROTON is the final computer deadtime correction we must evaluate. It takes into account the possibility that a background (or "singles") proton may arrive in the - 40 ns window between a coincidence electron and proton, thereby blocking the coincidence 100 - - 2000 - IIIIIIIII . I I I I I I I - 1 I I L 1- 1 I I I I , I L . . . . I I I I I I I I I I 1500 Cn D" 1000 0 U 500 0 . 0 I I I I , I I I I I I 7 I 20 10 30 40 TDC,, (ns) Figure 32: Example of raw coincidence timing spectrum measured by TDC8. This TDC is started by PROTON and stopped by EL20 (delayed) - inverse to the sequence which formed the COIN trigger. Consequently, larger time differences between electron and proton produce smaller values in TDC8- The peak at 35 ns is caused by EL20 signals overlapping with PROTON signals at the openng of the coincidence gate. The counts at 0 ns are due to TDC underflows and were excluded from the analysis. 101 proton. Note that tracks are certainly recorded for events within several tens of ns of the trigger which starts the TDC's and ADC's attached to the detectors - in other words, the tracking can certainly "see" several particles arriving within say 20 ns of each other. However, in the "one-track" mode, a group of equivalent tracks is reduced to a single choice by taking the track with the earliest time. Coincidence protons are then truly blocked by earlier singles. This choice is motivated in Section 32.1, but is also convenient for the evaluation of the proton blocking correction which follows. The central observation to make is that only singles protons can cause deadtime over and above what has already been corrected for: no matter how high the coincidence rate, K`P," has already corrected for unanalyzed coincidences. Let us use r,, r, , and r to refer to the coincidence, electron singles, and proton singles rates respectively; "singles" here refers to non-coincidence events only. Further, let t, refer to the effective - 40 ns window in which a PROTON may arrive before the coincidence timing peak and still form a COIN event. There are 4 possibilities for what causes a coincidence trigger, and each is restricted to a certain part of the timing spectrum: a coincidence electron and proton (ep) will fall in the peak, a coincidence electron and single proton (e p.) will fall before the peak (p. must come before p) a single electron and coincidence proton e,) falls after the peak, while a pair of singles (e. p,,) may fall anywhere in time. Thus only events involving a p., fall before the timing peak. The total number of events in this region, labelled NA, is simply the number of spills with an electron event (N,) times the probability of a p. arriving withing the t, coincidence window: NA = 1 _ e-rptc). (3.22) The number of events in the coincidence peak (N,) is the number of electron triggers IV, times the probability that e arrived before e, times the probability that no p. occurred 102 in the t, window: N, = N r, (3.23) e-rp,, tc r,, + r, The number we are interested in is the true number of coincidences, n, =- rT. From the previous equations one obtains nc == N n, + n, C Here, Nc is te N, (3.24) 1 N A N,- number of coincidences extracted from the corrected coincidence timing spectrum by random subtraction, and so the remainder of the expression is the net computer deadtime correction. The first term , n,,,N,+nc is just the total number of electrons over those giving triggers - i.e. the factor Kc"Pl .6 from Section 36.2. The final term is our proton blocking correction: K coinPROTON (3.25) N N, NA is the number of events in front of the timing peak and can be accurately determined without recourse to the value of tc by simply summing that portion of the ctime spectrum. N, is the number of electron events, but as NA is determined from the final coincidence timing spectrum, any restrictions the analysis places on electrons must be included in this number. In fact, the 16 tracking places a Cerenkov cut on every candidate electron and so N, is better estimated by the CAB40-V scaler than by the larger number E40-V. The CAB rates also involve shower counter cuts PA Low and PB, and though they are not very restrictive, there are no hard shower counter cuts imposed in the analysis. CAB40-V will then tend to somewhat underestimate N. Fortunately, however, CAB40-V and EL40-V record very similar values at exactly those kinematics where K ci,,PRITOI is large - at low Q and forward angles. At the higher Q, the increased pion rate in the 16 GeV/c 103 causes a greater discrepancy between the two scaler rates, but the background rate of proton singles in the eV/c is much lower. The net result is that the difference in correction factors computed using the two extreme values for N, is at most 1%. 3.6.3 Proton Absorption The detected proton rate is reduced not only by final state interactions with the residual nucleus, but also by similar interactions with materials encountered on the proton's path to the detectors. Information on the cross-section for nucleon scattering in various materials is available in ref. [5]. The total cross-section is composed of three components: (3.26) O'total ::--::O'elastic + Oquasielastic + Uinelastic- Of these, only the inelastic portion (which will cause at least m = 140 MeV of energy loss) is certain to remove a proton from the experimental acceptance. The elastic and quasielastic cross-sections are strongly peaked at small scattering angles at these energies, and may deflect a proton only minimally. Consequently, extreme upper and lower estimates of the percentage of protons lost may be obtained using O't,,t,,land Uinelastic re- spectively. The mean free path A is a convenient quantity to use for such calculations, as the classical probability of transmission through a material of thickness t is given simply by t Ptrammission = e In unitsof (3.27) g CM-2, A = 1.661A (3.28) 01 where A is the nuclear mass of the material in amu and is in mb. From Equation 3.27, we see that the total transmission probability through a sequence of materials is 104 simply e- X. The resulting correction factors are listed in Table 3 , after averaging between the upper and lower estimates given by AT and Al. The difference between the two estimates provides a systematic uncertainty of ± 2. consistently arger at Q = Note that the absorption is and 68 (GeV/c)'; this is due to the presence of gas in the 8 GeV/c Cerenkov counter, which was filled only at these settings. These results were verified for hydrogen at all Q by comparing the electron singles rate in the 16 GeV/c with the rate of coincidence protons detected by GeV/c. Ap- propriate corrections were made to the proton rate (e.g. for coincidence deadtime and identification efficiency). By placing cuts on the electron kinematics, one can ensure that the elastic proton cone corresponding to these electrons lies entirely inside the GeV/c, and so the probability of proton absorption is directly measured. The comparison is shown in Figure 33, and demonstrates that the absorption is understood to within the systematic error of the data. 105 ............ I................. . . . 1.00 I I I I . . . . . . . . . I . . . . . . . . . I I r. 0 0.95 .0.4 rn U) . P.4 - - - - - - -.- - - - - 41 5V) 0 - - - - - - - II ------------ 0 0.90 $.4 E- - - - - - - -0 0 0 0 0.8 1.4 .4.) 04 0.80 I I 0 2 4 Q2 (GeV/c 6 8 )2 Figure 3-3: Proton absorption as a function of Q1 40]. The dotted lines represent the results of a calculation using the total (lower) and inelastic (upper) cross-sections of the proton in the various materials encountered. The data point represent the ratio of (ee'p) coincidences to (ee') singles measured using a hydrogen target. 106 Table 31: Correction factors applied to the data. K comp,1.6 K comp,2/3 , and K compPROTON are the electronic and computer deadtimes discussed in Section 36.2. Note that these corrections were computed for each kinematic setting using the formulae given in that section; ranges of these values are given here for reference. Target 1H 2H 12C 56Fe 197Au Q2 I 3 5 6.8 I 3 5 6.8 I 3 5 6.8 I 3 5 6.8 1 3 5 6.8 I NOTRK1.6 p abs. 1.00 1.00 1.00 1.00 1.01 1.01 1.01 1.02 1.01 1.02 1.03 1.03 1.01 1.02 1.03 1.03 1.01 1.02 1.03 1.03 1.09 1.09 1.11 1.11 1.09 1.09 1.14 1.14 1.09 1.09 1.13 1.13 1.08 1.08 1.11 1.11 1.08 1.08 1.11 1.11 def f 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 1.015 KcOmP,1 .6 1.040 1.003 1.003 1.003 1.204 - 1436 1.018 1.014 1.004 1.061 - 1153 1.016 - 1023 1.014 - 1016 1.008 1.075 - 1108 1.016 -1.022 1.006 - 1.012 1.003 - 1.006 1.037 - 1.051 1.010 -1.012 1.013 1.009 107 Kcomp,2/3 KCOMPIMIT117 1.039 1.041 1.039 1.021 1.01 - 1095 1.000 1.099 1.002 1.001 1.001 1.002 1.002 1.001 1.001 1.001 1.001 1.001 1.001 1.002 1.002 1.002 1.001 1.001 1.000 1.007 - 1167 1.004 1.015 1.008 1.009 - 1027 1.003 - 1.008 1.01 - 1.068 1.010 1.019 1.013 - 1.026 1.006 - 1.014 1.00 - 1.007 1.005 1.009 - 1023 1.004 - 1012 1.010 1.005 Chapter 4 Description of the Experimental Simulation Each of the two forms of results derived from NE18 (spectral functions and transparencies) requires a Monte Carlo model of the experiment. As will be seen in the next chapter, the extraction of spectral functions requires a model of the coincidence phase space of the experiment - the probablility of obtaining a given E, and p,. As explained in the introduction, the transparency is defined to be the ratio of measured to PWIA yield, and so in this case a full model of the PWIA cross-section described by equation 17 is required. Since radiative corrections can be modeled accurately using such a Monte Carlo calculation, one is also able to compare the extracted spectral functions with the Monte Carlo results, thereby compensating for the distortions to S(Enpn) caused by radiation and obtaining a good estimate of the success of the PWIA model in this context. Each component of the PWIA calculation is discussed in detail below. 108 4.1 Overview The Monte Carlo program, named SIMULATE, randomly generated the momenta and angles of the scattered electron and proton vectors (i.e. the 6 quantities in terms of which the differential cross-section is defined) with a flat distribution over limits calculated to exceed the experimental acceptance. The energy and position of the incident electron at the target were also generated randomly, to match the energy and spatial spread of the beam, and the beam energy was corrected for ionization losses in the target. With a basic event at-,the scattering vertex now determined, the possibility that any or all of the particles emitted real or virtual photons was modelled and the particle vectors were adjusted accordingly (see Section 45). The scattered electron and proton vectors were then transported through the target, applying ionization losses and a multiple scattering distribution, and subsequently transported through the spectrometers. "Single arm" Monte Carlo models of the optics, apertures, and interfering materials of the spectrometers were employed and are described in Section 44. Both forward and backward sets of matrix elements were used, to simulate the optical resolution of the magnetic systems. Once the particle vectors were reconstructed back to target, they were corrected to the scattering vertex using the mean energy loss calculations employed in the analysis, and E and p, were determined. The successful events were stored in histograms, with each event being assigned a weight of Ko,,pS (1-6h.rd) 1 Wg,.,,. The factor ( - h,,d)-' is a correction for radiative diagrams involving hard virtual photons, and is explained in Section 45. The "generation weight" (Wg,,,) comes from the following source. To increase computer speed, the limits in which event quantities are generated can be refined once partial information about an event is known. These refinements are based on the acceptance limits of the spectrometers, the cuts imposed on reconstructed E, p, and particle vectors, and the range in E, (at the vertex) over which the spectral function is defined. These 109 refined limits are especially important in the generation of radiation. For example, to take into account the possibility that a scattered electron "radiated into" the 16 GeV/c momentum acceptance from a higher momentum, one must use generation limits in 1.6 which are much wider than the acceptance. However, once the electron's momentum has been generated, one can determine the range of photon energy required to produce a successful event. The generation weight reduces the event weight to compensate for the restricted limits employed. Finally, the results histograms were normalized so that the number of events in each bin would correspond to the number of counts expected from the experiment. The results were thus multiplied by L(AF_'AQ"'AE'A0P')g_Ngen where C is the experimental luminosity, and the other terms refer to the phase space volume and total number of events used by the generation. Each histogram bin was assigned an inverse fractional error equal to the square root of the number of Monte Carlo events contributing to the bin. 4.2 Off-Shell Prescription for p As discussed in Section 12, the deForest prescription 3] for the fundamental cross-section for quasielastic electron scattering from a bound nucleon is used. His 0',,, prescription for the nucleon structure functions is used in our final results. However, the effect of the uncertainty in o,P was estimated by performing additional calculations, using 0',,2 as well as a simple on-shell prescription obtained by replacing the off-shell kinematic variables in o,,, by their on-shell values. The net PWIA cross-section was found to change by less than 2. The structure functions are computed in terms of the form factors the free proton. We used the dipole form for the proton electric form factor GP E and the parametrization of ari and Kriimpelmann 411 for the proton magnetic form factor GP 110 These forms were determined by experiment NEII [1] to provide the best description of inclusive electron scattering data up to Q of 4.3 (GeV/ C)2. Model Spectral ]Functions The model spectral functions used in the PWIA calculation are based on the Independent Particle Shell Model (IPSM) - i.e. the nucleus is described by a sum over nucleons occupying distinct shells. Furthermore, we assume the customary first-order approximation that the spectral function can be factored into separate functions p(p,,) and CE,). S(E.,p. = 2' + 1) 1: Pn,1j(Pm),Cn,1j(Em) (4.1) For hydrogen, of course, the spectral function can be described trivially by delta functions around Em = and p, = . In practice, the Monte Carlo procedure is altered to generate only the electron angles; all other kinematic quantities can then be computed. The nucleon momentum distributions p(p,) were computed by the program DWEEPY [42], obtained from NIKHEF-K, which solves the Schr6dinger equation in an optical model potential. (It should be noted that DWEEPY provides many additional op- tions and output modes, including extensive calculations of distorted wave effects CDWIA) and full five-fold cross-section computation.) In the case of deuterium, the wellestablished Bonn potential was used 431. For the other nuclei a sum of Woods-Saxon and Coulomb potentials was employed 4: Yom = Vof(r, Ro,ao) + V h MC 22 df -- I s + V(r). r dr (4.2) The Woods-Saxon potential describes the average nucleon-nucleon interaction and is composed of a volume term depending only on the nucleon's radial position r and a III spin-orbit term which takes into account the coupling between the nucleon's spin and angular momentum. The radial shape of each term is described by the Woods-Saxon (or two-parameter Fermi) form: f (r, R, a) I+e (4.3) aR The radius parameter R is in practice replaced by another parameter ro (or r,,), defined so that the bulk of its dependence on the number of nucleons in the residual system is removed: R = ro(A - (4.4) 1)1/3. The Coulomb term is simply an expression for the potential produced by a uniform spherical charge distribution of radius RC: (Z - I)e Vc (r = where RC =_rA11'. 4rcoRc 2 3 r2 2 2RC (4-5) 2 The optical model provides a strictly local potential, while the true potential is in general non-local, depending on the value of the nuclear wave function at other locations. This complication is typically approximated by modifying the nuclear wavefunction computed from the local potential. One such prescription is that of Perey [44], involving a range parameter V)NL(r) V)LM (4.6) [1 + 2M02VOM(r)]1/2 The set of parameters I (O I ro, ao, Vso, rso, a.9o, I provides a complete description of the nucleon wavefunction in a particular shell; the binding energy EB of the shell can thus be computed from these values. One can alternatively choose to provide DWEEPY with a value of EB as input, and have it compute the well depth V. This option was only used when no information on could be found. If both parameters were available, o 112 was used in p(p,) while EB was used inC(E,,,). The binding energy distributionsC(E,,,) were described using the Lorentzian (or BreitWigner) shape associated with an isolated resonance: L(E) where EB r/2 7r(E - EB )2 (4.7) (IF/2)21 refers to the central binding energy of the shell. The minimum proton sepa- ration energy Emin of a nucleus is given by the mass difference MA-, 3 Mp - MA. We incorporate the lack of spectral function strength at binding energies lower than this by imposing a cutoff onC(E) at E.,min and renormalizing the Lorentzian to 1. Values of E.min for nuclei relevant to the discussion below are listed in Table 4. The width parameter r was either taken from fits to JOWQ2data (as explained below), or from the formula of Brown and Ro 45] f'(E) (24MeV (E - EF)2 (4-8) (50OMeV2) + (E - EF)2 with E taken to be the central binding energy EB of each shell, and E, the Fermi energy. This formula is a parametrization of the widths of single particle states, used to fit data on A < 8 nuclei. In Ref. 46], however, (ee'p) data on 201Pb demonstrated that the expression also works well for heavy nuclei. An alternative where the fixed width version of L(E) was also used, in Equation 47 was replaced with a running width I'(E) from Equation 48. The difference in results between the two forms of L(E) was found to be less than 2 for all targets and kinematics. The Woods-Saxon parameters for 2C weretaken from Saclay12C(ee'p) data 47 at Q 2 of 016 (GeV/C)2, and are listed in Table 42. Ref. 4 also contains a convenient summary of the Saclay experiment. No non-locality correction was applied to the wavefunctions used to fit the parameters, and so we also do not apply it. The binding energy for the Is shell was taken to be the centroid energy of 38.1 MeV reported in 4 113 The Table 41: Minimum proton separation energies for nuclei discussed in the text. I Nucleus.1 E 3mi.n(MeV 12C 15.96 56Fe 10.18 197 Ni Au 8.17 5.78 208 Pb 8.01 58 I Ip shell energy was set at 16.2 MeV, from a combination of various experimental results detailed in Table X of 4 and an inspection of the E spectrum found at Saclay. The shell widths ' were also determined from an examination of this spectrum. The 16 Fe model parameters were taken from (ee'p) data on the nearby nucleu 5Ni (also from ref. 47]), and are listed in Table 43. The Woods-Saxon parameters were obtained from Table VII of 4 and again the non-locality correction was not applied. However, p(p,,,) extracted from NE18 data at Q = (GeV/C)2 showed a narrower distribution than this model. A better fit was found by simultaneously widening the radius parameters ro and r,,, from 126 fm to 13 fm, and applying the non-locality correction with range parameter = 5. The results presented in the next chapter are calculated using this best-fit model, but the low-energy model was also used, to obtain a model dependent uncertainty. For L(E), the shell energies were taken from the same source (Table X of 4. However, the experimental data on the deepest shells (15 and p) are quite uncertain, and a second set of values, obtained from the Hartree-Fock calculation of 48], we also used. These theoretical values were found to provide better agreement with our data. All of these binding energies were increased by 20 MeV to correct for the difference in separation energy between 18 Ni and 16 Fe (see Table 4 1). The shell widths were calculated from Equation 48, using a Fermi energy of MeV. The model spectral function for "'Au (see Table 44) was obtained using the 'O'Pb(ee'p) 114 data of ref. 46]. Fits of ro were only performed for shells near the Fermi surface and are found in Table 45 of 46]; discussions with a collaborator on the experiment 49] revealed that the version of the DWIA corrections presented in column (b) of the table is the best one to use. For the deeper lying shells, the typical value for heavy nuclei of ro = 118 fm was used. Or Q = (eV/c)' data on (p,) were again found to be narrower than the model. The radius parameters were consequently increased by a factor of 1.1 in our best fit model. Potential well depths Vowere not fit by the 108 Pb(ee'p) experiment, and so the shell energies were provided instead as input to the Woods-Saxon calculation. These are found in Table 49 of 46] for all but the deepest shells, which were beyond the experiment's detection range. Energies for these shells were taken from theory 48] A second set ws determined from the collection of separation energy measurements presented in ref. [50] (and also in Figure 117 of ref 4 These measurements extend only to A < 60; values at A = 197 were obtained by extrapolating the trends in the energies for each shell. These extrapolations are certainly accurate to no more than 20%, but as they affect-,less than 10% of the protons in "'Au, their effect on the results is small compared with other uncertainties. The 22 MeV difference between E"' 211 for 97 Au and Pb was subtracted from all energies. The remaining Woods-Saxon parameters are shell-independent, and taken from the discussion of the optical potential on p.15 of [46]. The widths for the four shells most near the Fermi surface were taken from Table 4.9 of 46]. The rest were not obtained in that experiment, and so we computed them from Equation 48, using a Fermi energy of 44 MeV. This value of EFwas that used by Quint corrected for the 22 MeV mentioned above) in his calculations using the Brown and Rho formula- as mentioned earlier, he found that the formula agreed well with the measured widths. 115 -Table 42: Model spectral function parameters for 12C Woods-Saxon parameters for p(p,), from data of 47]. No non-locality correction was applied. 11 V0 ro mev fm 6 55 1.36 1.36 ao she 1sj/2 IP3/2 qo Lorentzian parameters for CE,), rso aso MeV fm 0.55 0.55 9 1.36 rc fm 0.55 1'3 1.3 from the same source. The Fermi energy used was EF 15.96 MeV. EB r, MeV MeV IS1/2 38 20 IP3/2 16.2 5 shell 4.3.1 Correlation Corrections The IPSM provides a good approximation to the nuclear wave function, but by construction fails to account for any correlation between nucleons. In fact these correlations have a measurable effect on the spectral function. They are primarily due to the strongly repulsive nature of the nucleon-nucleon interaction at short distances, and this repulsion tends to accelerate the affected nucleons to higher momenta. Our approach to this problem is as follows: We assume that strength shifted to higher p, by these shortrange correlations only becomes comparable to the IPSM strength well past the Fermi momentum, and so beyond the detection range of our experiment. In other words, we assume that the IPSM correctly describes the shape of the spectral function we observe but should be corrected for the integrated correlation strength outside our phase space. This assumption is supported by both nuclear matter calculations [51] 521 and spectral function measurements made at low Q [6]. Ultimately, of course, the test of our assumption will be the agreement between the model and the data. 116 Table 43: Model spectral function parameters for 56Fe Woods-Saxon parameters for p(p,), from data of ref. 471 on "Ni. To improve the fit to NE18 data, ro was widened to 13 fm and the Perey non-locality correction was applied with range parameter = 5. Calculations using ro = 126 and no non-locality correction (the fit determined in ref. 47]) were also performed, and used to determine a model-dependent uncertainty. shell VO ro ao MeV fm 1S1/2 IP3/2,1/2 ld5/2,3/2 2s,/2 1f7/2 80.7 69.0 58.2 52.7 52.7 1.30 1.30 1.30 1.30 1.30 0.60 0.60 0.60 0.60 0.60 Vso r,,, MeV fm a., 40 23.5 1.30 1.30 0.60 0.60 13.8 1.30 0.60 rc fm 1.3 1.3 1.3 1.3 1.3 Lorentzian parameters for CE,,). The 1d, 2s, and f binding energies are from the Hartree-Fock calculation of ref. [481-those for s and Ip are from ref. 471. As both of these sources are descriptions of "Ni, the binding energies were corrected for 20 MeV difference in proton separation energy between "Ni and "Fe. The energy widths are given by Equation 48, with E = MeV. Calculations were also done using the Is and Ip binding energies from ref. 47]; these are indicated in parentheses. 117 Table 4 Spectral function parameters for 197 Au Parameters for p(p,) and (E), from data of ref. 461 on 211 Pb. Note that EB rather than is provided to the Woods-Saxon calculation, and that these values are also used to determine LE). Also, as all EB values are for 208 Pb, they are corrected form the 2.2 MeV difference in proton separation energy between 208 Pb and 197 Au. The Perey non-locality correction was applied, with range parameter = 5. To improve the fit to NE18 data, ro was widened by a factor of 1.11, but calculations using the original ro values were also performed and used to determine a model-dependent uncertainty. Two versions of EB for the deep shells are also provided. The first values are from the Hartree-Fock calculation of ref. 48], while those in parentheses are derived from the data of ref. [50). shell 181/2 IP3/2,1/2 1d5/2,3/2 2sl/2 1f7/2,5/2 2P3/2,1/2 199/2,7/2 2d5/2 IhjI/2 2d3/2 EB F MeV 'ro fm ao MeV 46 65) 41 (50) 32 40) 28 18 21) 17 19) 13 17) 11 1.31 1.31 1.31 1.31 22.9 17.2 12.5 8.3 7.7 6.2 9.4 7.9 6.0 3.7 4.0 3 1.31 1.31 1.27 1.32 1.29 1.36 118 Vso MeV r,, fm ajo rc fm 0.65 0.65 0.65 0.65 6 6 1.15 1.15 0.65 0.65 1.2 1.2 1.2 i.2 0.65 0.65 0.65 0.65 0.65 0.65 6 6 6 6 6 6 1.15 1.15 1.15 1.15 1.15 1.15 0.65 0.65 0.65 0.65 0.65 0.65 1.2 1.2 1.2 1.2 1.2 1.2 Table 45: Corrections applied to the IPSM spectral functions to correct for strength shifted outside our detection range by short-range correlations. I Nucleus 12C Fe "Au I F,,orrel I 1.1 ± 003 1.26 ± 008 1.32 ± 008 We divide the IPSM spectral function by a correlation factor SjpsmdpdE F,orrel Forrel: (4.9) Scorreld'pdE The overall normalization f SdpdE this normalization. = Z is maintained since Sorrel is also defined with Two spectral functions Scorrel were used for carbon and are from ref. 53] and [51]. The latter is a calculation of the correlated spectral function for the nearby nucleus `0 using the Brueckner theory of finite nuclei. For the heavier nuclei iron and gold, we use correlated nuclear matter spectral functions, corrected for finite nucleus effects; these are from ref. 52] and 541. The correlation corrections applied to the IPSM spectral function are listed in Table 45. The indicated errors account for the different values produced by the various models. Note that, as mentioned earlier, we use several IPSM models to obtain a model dependent uncertainty in our results, and the value Of correlmust be recomputed for each Sjpsm. The values listed in the table apply to the best-fit IPSM models used to generate our final results. 4.4 Spectrometer Models The single arm spectrometer models employed sets of forward and reverse matrices to transport particles between the pivot point and the focal plane. The reconstruction 119 matrices derived from the data and used in the analysis (see Section 33) could not be used in this context for two reasons. First, inverting a set of > second order matrices cannot be done analytically, and if done numerically is not guaranteed to produce reliable results. Second, the magnetic field needs to be modelled in stages so that magnet apertures can be checked at intermediate locations in the spectrometers. Consequently, the spectrometer models used in the Monte Carlo calculation were derived from TRANSPORT decks. The following criteria were used to evaluate the models: * The forward matrix elements should reproduce the data distributions at the hut (focal plane). * The reverse matrix elements should reproduce the data distributions at the target. * The elastic cross-section determined by dividing (normalized) experimental counts by the spectrometer acceptance, as calculated by the Monte Carlo, should have the correct magnitude and be independent of the cuts applied. * The forward and reverse matrix elements should invert each other as nearly as possible. As described below, the GeV/c TRANSPORT model was modified to meet these cri- teria. The spectrometer Monte Carlos proceeded by transporting particles through the spectrometer in steps, stopping at each aperture to check the acceptance and at each intervening material to compute energy losses and multiple scattering. Energy straggling was calculated using equations 31 and 32. Multiple scattering was simulated via Gaussian distributions in each of the planar angles dx and dy, each with a standard deviation (in radians) of [5] O'dxdy '--- 13.6 z vft7d(I + 0.088 1091o(tad)) Po 120 (4.10) Here, z is particle charge, p is particle momentum in MeV/c, and tad is the thickness of the material in -units of radiation length. Once transported to focal plane, particle tracks were required. to intersect the fiducial detector volume. The tracks were also adjusted for multiple scattering in the detector materials before the wire chambers, and for the resolution of track reconstruction due to wire spacing. Finally, the adjusted tracks at focal plane were reconstructed to target. The 16 GeV/c optical model was taken from the work of ref. [55]. This model was developed for, a previous experiment at SLAC which used the same configuration of the 1.6 GeV/c as NE18. The 16 dipole is modeled in segments, producing sets of forward matrix elements. The GeV/c model was based on a second-order TRANSPORT model developed at SLAC 561 for the large acceptance tune used by NE18. Careful assessment of the uncertainties in measured magnet currents was made, and their values were adjusted within these limits to provide optimal forward-backward agreement. The forward model was found to reproduce the data at the focal plane but cut dependences were noticed in the E) and D coordinates. In D, the Monte Carlo distribution was found to be narrower than the data. A solution to the problem, which reduced the cut dependence of the elastic (ee'p') cross-section to < 2, was found by multiplying the matrix elements by a factor of 1025. Some sacrifice in forward-backward resolution was necessary with this change, but this sacrifice is not critical; The primary problem induced by such a mismatch is the incorrect assignment of vertex quantities to reconstructed target quantities in the PWIA calculation. As the out-of-plane cross-section varies slowly and multiple scattering smears the resolution, the indicated compromise is entirely reasonable. In 0, it was concluded that the trouble was related to the third-order matrix elements needed in the data reconstruction (see Section 33). Replacing the 121 matrix entirely with the elements from the data. set produced the desired result of reducing cut dependences again to < 2%. The final transparency results in general show a higher cut dependence of 5%. The uncertainties in the GeV/c spectrometer model are largely responsible for this, and we invoke a 5% systematic uncertainty at all targets and kinematic settings to take into account this lack of understanding. 4.5 Radiative Corrections The PWIA model described thus far involves only one Feynman amplitude, for the exchange of a single virtual photon between the electron and struck proton. However, one must also consider the possibility of emitting additional photons, both real and virtual. The emission of real photons is referred to by the miliar term "Bremsstrahlung" or "braking radiation". These photons are emitted when the charged particles involved in the reaction are accelerated by the fields of either the nucleus involved in the primary hard scattering ("internal radiation"), or by the other nuclei encountered by the incoming/outgoing particles as they travel through intervening material ("external radiation"). The emission of real photons causes a discrepancy between the detected particles' momenta and their actual momenta at the scattering vertex, and so cause distortions in the extracted experimental spectra. Conversely, amplitudes involving the emission of additional virtual photons affect only the magnitude of the measured cross-section. As will be demonstrated below, the Bremsstrahlung cross-section is strongly dependent on W. In the search for a Q-dependent signature like that of colour transparency, then, radiative effects must be computed precisely. In the following discussion, we will use to represent the four-vector of an emitted real photon. This appears in the energy-momentum conservation relation as an additional 122 four-momentum in the final state: PA = P' w+ * (4.11) All of these variables are four-momenta, representing respectively the initial electron, the inital target nucleus, the scattered electron, the knockout proton, the emitted photon, and the recoiling (A- 1) system (possibly in an excited state, as indicated by the asterisk) If one now denotes the values one measures for the missing momentum and energy by PM and , and their actual vertex values (in the absence of radiation) by p. and Em, one obtains pm = pw-q=Pm+w Em + Tre = E E- E - M - wo = km + Tec WO, (4.12) and so PM = PM-W E. = Em + T Em - Trec+ W0 w0 Note that the measured value of the recoil kinetic energy, (4.13) cj depends on the measured missing momentum and so is also distorted by Bremsstrahlung photons. However, the contribution of Trec to the missing energy is, in general, small (and non-existent in the case of elastic ep scattering). This approximation is not used in the radiative description, but merely serves to illustrate the overall effect of radiation on a measured (Em,pm) distribution: the real photons produce long "tails" which, at very high photon energy (W > Em, pm), are described by the relation Em '-- ,,, 123 wo. Elastic ep scattering I $ 0 a I _IgSbN_ -------------------------- - - - - - - - - - -___ ------Figure 41: Distribution of counts in E, and p, for (ee'p) from hydrogen at Q = (GeV/c)', demonstrating the existence of "tails" due to Bremsstrahlung radiation. The E, axis runs in the bottom-right direction, from 25 to 125 MeV in bins of 25 MeV; the p, axis runs towards bottom-left, from 160 to 160 MeV/c in bins of MeV/c. provides a clear demonstration of these tails, since in the absence of radiation, all strength is localized at E = ,, = (see Figure 41). The coincidence variables E, and p thus provide a natural coordinate system in which to evaluate radiative effects. By contrast, radiative corrections have generally been calculated in the framework of inclusive (ee') experiments - in terms of their effect on the measured electron energy c = P. The effect of radiation on this quantity depends on the direction of the emitted photon: Consider elastic scattering, with differential cross-section dQi d The reaction amplitude is thus fixed by the direction of the scattered electron (and, of course, the incoming electron energy). If we treat this direction as fixed, the radiation of a photon parallel to k simply decreases the energy E' by the 124 photon energywo. if, however,the photon direction is parallel to the incoming electron, E' is affected by an amount that depends on the electron scattering angle. (Note that the scattered proton vector is also affected). Thus, when one comes to evaluate the total probability of emitting radiation that affects by less than some cutoff energy AE, one has to perform integrals over photon energy and direction with interdependent integration limits. In the case of coincidence scattering, independent integrals can be performed as the measurement of both scattered particles enables one to select a more "natural" choice of variables. The formalism described in this section 571 is based on the work of Mo and Tsai [58] 59], which has provided the standard radiative corrections prescription for three decades of inclusive electron scattering experiments. The basic formulae of Mo and Tsai have been reevaluated in a coincidence framework: one can no longer integrate over all final states of the scattered proton as in (ee') measurements, but must calculate the radiative effect on both the scattered electron and proton. The resulting distributions are then included in the event generation of our PWIA Monte Carlo and folded with the experimental detection range in k' and p as described earlier. 4.5.1 First Order Internal Bremsstrahlung The probability for radiating a single Bremsstrahlung photon is represented by the four Feynman diagrams of Figure 42. Since each of these diagrams involves the same final state, the amplitudes must be summed coherently: da d3kld3W "IIM', + M.f + MP, Mpf 12. (4.14) These four matrix elements refer to the emission of a photon by the incident electron, scattered electron, incident proton, and scattered proton respectively. To evaluate them, 125 one requires a knowledge of the coupling of the electron and proton to the photon. The electron coupling is given exactly by QED and is specified by the electron current J,'(q = eie(k + q)-ylu,(k). Here, e = (4.15) r4ra is the electron charge and u, is the electron spinor, normalized to U,(k)u,(k = 2m (m is the electron mass). The proton-photon coupling is complicated by the fact that the proton is in general bound and off-shell, and the description of such a proton is only approximately known (see Section 12). Consequently, we neglect these effects and assume elastic scattering from an on-shell proton: JI(q = -eiip(p + q)P(q)up(p). (4.16) The deviation of the proton from a point particle is described by I" (q = F (q'), I + - I 2M F2(q')Zo,`q,, (4.17) using the free proton form factors. Again, the proton spinor is normalized to the proton mass: p(k)up(k = 2M. In support of this neglect of off-shell effects, one can point out that the overall contribution to the radiation from the proton arms is expected to be small. Using these couplings, one obtains the following expressions for the first-order Bremsstrahlung matrix elements: i,"(k Me = zue(lk'),A (k - Mf = 1ii,(k')e-y',E, A4p = 'iip(pp(q) - w,) + 2 e-"E'U'(k) q2 - P2 _U W)2 - MI (kl+ + w,) + m L;)2 -t u, (k) M2 e2 (p')F,.(q)up(p) UP(p')rjq)up(p) e2 (p - W)2 _)F'(W)'E'UP(P) M2 q 126 2- A 2U ,(k')7.u. (k) M'f M'ei I _F + Mpf Figure 42: Feynman diagrams contributing to first order Bremsstrahlung radiation cross-section. M Pf up(p) LI (W),E iy"(P' w)- + (p + + O)2 M2M e q - Y2 Here, c, is the polarization of the Bremsstrahlung photon, q = p - p is the momentum transfered to the proton if the electron emits the photon, and q = k- k is the momentum transfered to the proton if the proton emits the photon. the photon is a parameter representing ass, which will ultimately be taken to . The single hoton emission cross-section can be calculated from these expressions, with no further approximations. However, the formulae simplify greatly in the limit that the photon energy wo is much less than the momenta of the initial and final state fermions. In this case, the basic one-photon exchange amplitude MM ep factorizes from the Bremsstrahlung amplitudes, giving M = eM ep 127 w k k = eM (1) ep M'f ,E k' k' Mpi = -Ov I ep1) Mpf = -(, -em(l) ep P (4.19) W-P, This limit is referred to as the soft photon approximation (SPA); it can be seen to be reasonable from the distinctive energy dependence of the emission amplitudes. Part of this approximation is the use of the elastic (unradiated) values of the fermion momenta k, k p, and p in the above expressions. These elastic values are also used in the evaluation of the one-photon exchange amplitude, M(1) ep = i,(k')-y"u,(k)_ e2 q2- 9 2 ii p') r, (q) up p). (4.20) The resulting total cross-section for single-photon Bremsstrahlung is thus given by do, dQd3 w d dQ, a ep 47r 2WO k' w k' P w p k w k p w p 2 (4.21) For later convenience, we write this as a product of photon energy and angle distributions, do, dQdQYdwO du dQ, A(Q,) P (4.22) W0 A (Q_ =- -- -+ where aw 02 k' 4r2 w k PI w p k w k p -p 2 (4.23 depends only on the photon direction L. Integrating Equation 421 over photon angle and energy, one obtains the cross-section for emitting a photon of energy less than AE: do, dQ, (WO< A E AE = Jo do, d3WdQd 3W 128 do, dQe 2a) 0 (pi) E)(pj) B (pi, pj, A E), (4.24) ep where AE B (pi, pJ, A E) 3W Pi d 87r2w, P Pj pi) P pi), (4.25) Here, two pieces of convenient notation have been introduced. pi for i = 1,... to represent the four fermion momenta k, k, p, p in turn; the constants 4 is used (pi) denote the signs accompanyingeach term, 0(k = 0(p' = -1 and O(P = 0(p = 1. This integral can be evaluated using the expression w k = w(ko - Jklcos 0), as well as introducing (4.26) a new variable x as indicated in Equation P., = Xp + (I 111.19) of Tsai 59]: XPj. (4.27) One then obtains B (pi, p, A E) P'2rpj Pi 01 dx IAE 0 'p I 2r dx o 2 PX 0 Wo P2W2+ X In AE 0 PX 0 + P2 (P 0)2 X 2 2p2X In P P 2 +PX - 1PX1In PX- 1P.1 + In _2pX POX + 1P.1 21pxl We note that the sum Ej 0pi)0(pj)B(pj, 0 (4.28) POX+1P.1 pj, AE) is negative, making the total cross- section (and the angular distribution A(L)) positive. One observes that this expression contains two non-physical divergences: when the C4 photon mass" p -- 0 and when the energy cutoff AE --+ 0. Both of these are due to approximations made so far, and will be addressed in later sections. Before continuing, 129 however, it is worthwhile to try to evaluate the validity of the soft photon approximation. As mentioned above, the one photon Bremsstrahlung calculation can be computed without this approximation. Accordingly, the ratio of the full to the soft photon calculation is presented in table 46 for Q = I GeV/c)' and a variety of photon energies, and in table 4.7 for a photon energy of 100 MeV and a range of Q from I to 15 (GeV/c)'. Qual- itatively, one sees that the SPA improves at low photon energies and high momentum transfers, as expected. At Q = I GeV/c)', the discrepancy between the two calculations is less than 1% for photon energies less than 10 MeV, while for a photon energy of 100 MeV the discrepancy drops to 5% at Q = 9 GeV/c)'. The discrepancies are considerably higher at the other settings listed, however. Two effects are involved: the shape of the Bremsstrahlung energy spectrum, and the evaluation of the matrix elements using elastic (wo = 0) particle vectors (i.e. neglecting the difference between q and q in Equation 418). In an attempt to separate these effects, table 46 also contains the SPA to full ratio using a point-like proton, i.e. a proton without form factors but with the correct magnetic moment. At Q = I GeV/c)', one sees that most of the discrepancy is removed along with the q-dependent form factors. To resolve this one must evaluate the cross-section using a value of q which is corrected for the effect of radiation. In other words, one must distinguish between photons emitted before and after the hard scattering, a task which is complicated by the interference terms between the Bremsstrahlung amplitudes j, Mf, Mpi, and Mpf. However, such a correction can be built into the calculation, as is explained later on. The maximal E range considered by NE18 is about 140 MeV, and so the wo = 100 MeV results in table 46 can be considered a worst-case scenario. Assuming that the correction to q at the hard scattering vertex can be accomplished, one is faced with a SPA inaccuracy of at most 2 the direction k and 7 for radiation in the direction P 130 for radiation in We point out in passing that Table 46: Ratio of single photon Bremsstrahlung cross-section calculated in the soft photon approximation to the full calculation, at Q = I GeV/c)'. Various photon energies wo are considered; the photon angle is taken to be in the direction of either the initial (i) or final (f) electron. The values in parentheses are the SPA/full ratios using a point-like proton in the calculations. I Wo (MeV I 1 10 100 200 i 1.0023 1.023 1.26 1.59 f 1.0002) 1.002) 1.02) 1.04) 09993 0.993 0.93 0.87 0-9993) 0-993) 0.93) 0.87) Table 47: Ratio of single photon Bremsstrahlung cross-section calculated in the soft photon approximation to the full calculation, for photon energy = MeV. Various momentum transfers Q are considered; the photon angle is taken to be in the direction of either the initial (i) or final (f) electron. IQ' (GeV/c) I 1 5 9 15 f 1.26 1.14 1.05 1.03 093 093 099 099 these percentage discrepancies are given in terms of the radiative corrections, which are themselves small; the effect of these discrepancies on the final cross-section is thus much less than the quoted percentages. 4.5.2 Virtual Photon Corrections One of the non-physical divergences observed in Equation 428 was found in the limit y ---+ 0. This is known as an "infrared divergence", and is a direct consequence of the fact that the one photon Bremsstrahlung cross-section is of order a 3 and that other diagrams of the same order have not been included yet. These are amplitudes for the emission of 131 two virtual photons, collectively referred to as M('). ep These must be summed coherently with MM which represents the same final state: M2 = IM(1)12+ M(2)tM(l) + M(l)tM(2) + O(C,4). ep ep ep ep ep ep (4.29) Figure 43 contains a summary of the second-order amplitudes. Unfortunately, several of these depend implicitly on the strong interaction via the poorly known proton current. The point of view advocated by Mo and Tsai and espoused here, is to include only those terms which do not unambiguously depend on the strong interaction. Certain amplitudes such as M(2.1) in the figure are calculated, but only infrared divergent terms necessary ep to cancel those from the Bremsstrahlung cross-section are kept; the rest are left buried in the electron-proton cross-section. It should be noted that other workers 60] have derived alternative expressions for the virtual radiative correction, by including some of the components left out by Mo and Tsai. However, the point to be made here is that the evaluation of MM includes the use of proton form factors extracted from previous data. The radiative corrections applied should thus be onsisted with whatever corrections were used in extracting these form factors. The standard prescription of Mo and Tsai is thus the appropriate choice [I]. The second order diagrams depicted in Figure 43 are grouped into three categories depending on their sensitivity to the strong interaction. We use the same evaluation of these amplitudes as Mo and Tsai, and restate them here. Also used is the notation K(pi, p = pi p 1 dx In 3 10 132 2 Pr p x2 2 1 IL (4-30) M (2-1) M (2.2) ep ep M 2.3) ep Figure 43: Feynman diagrams representing virtual photon corrections to one-photon exchange ep cross-section included here and in ref. 591. describing the form of the infrared divergent terms. Note that K(pi, pi) =In (4.31) Mil Y2 and that the IR divergent term of Equation 428 has this form. ./" the ep2. ) limit is the electron-photon vertex correction and is known exactly from QED. In Q2 > (which is well satisfied by momentum transfers in the GeV/c range), M2 one obtains a M(2-1) - ep M(2.2) ep 2ir - K(k, P) In(-Tn2) + -3In(-- q2 ) - 2 M(1) IL2 2 rn2 (4-32) ep represents the vacuum polarization correction, and contains contributions from both lepton and hadronic loops. Only the former are known unambiguously from QED, and contribute M ep2.2 = a E 6VP1 M ep (1) i i 133 (4-33) whereEi sumsover the different flavors of leptons with mass mi. Again applying the limit Q2 > m2 , one obtains bVP 5 - 37r 3 +In q (4-34) M?I As there are no IR divergent terms in the vacuum polarization amplitude, all contributions from the strong interaction are neglected. Finally, M(2.3) ep includes two-photon exchange and nucleon self-energy graphs, both of which depend intrinsically on the strong interaction. Only the IR divergent terms are used: M(2.3) - _a ep 2ir K (k, p) + K (V, p - K (V, p - K (k, p') - K(p, p') - In (M2) (4.35) 92 The total cross-section for emitting a photon with energy less than AE is now obtained by adding all of these terms to Equation 424. The dependence on the photon mass cancels as required, leaving do, 0 P < AE) dQ, doO) lep (I - bsoft(AE - 6haxd) i dQ, (4.36) where bsoft(,AE = 2a 0 (p 0 (p 13(pi, pA E) (4-37) and shard 2 - 3 4r ln(-q 2/M2) + 1 ir 2) bvp(q (4.38) Here, d7 Jp represents the one-photon exchange ep cross-section, bhwd i the contribudQ, tion from the second order virtual photon diagrams, and bsoft(AE) is due to one photon Bremsstrahlung. f(pi, pj, AE) is simply B(pi, pj, AE) of Equation 428 without the IR 134 divergent term. In order to seperate out the contribution of the proton we divide &,ft(AE) into three parts, bsot(AE = ee soft (AE) + bep soft (AE) + 6PP soft (AE). (4-39) 6ee soft is the electron Bremsstrahlung contribution, involving f3(k, k, AE), P(P, V, AE), and B(k, V, AE). bP soft includesthe electron-protoninterferenceterms B(k, p, AE), B(k, p', AE), B(k p, AE), and B(k , p, AE); while bsoftis entirely due to proton radiation and includes the remaining terms B(p, p, AE), P(p, p', AE), and B (p, p', AE). Table 48 contains values of these terms as well as bh,,,dat various kinematics. Note first that bh"rdis negative, and so causes a net increase in the total ep cross-section. Its magnitude is also small: less than 10% up to Q of 15 (GeV/c)'. The direct proton contribution (lowest Q2) to 8% (highest Q2) of the electron contribution soft . bee varies from 2% However, the electron- proton interference is about twice the size of the direct proton term, leading to a net 6-20% contribution of proton Bremsstrahlung. It is clear that proton radiation, though afflicted by strong interaction uncertainties, cannot be neglected at these energies. The evaluation of the functions f(pi, pj, AE) must be done numerically. Numerous Spence functions must be computed, where (X = x - In(Il - yj) dy. 10 (4.40) Y As an aside, the contributions of these functions turn out to be important only when their arguments are large (jxj > 1), and in this case an excellent approximation is provided by .D(x) 2 In 2(IXI). (4.41) The formulae simplify, however, in the "ultra-relativistic (UR) limit" where the momen- 135 Table 48: Values for the radiative correction functions 6, evaluated at various momentum transfers and for cutoff photon energies of 10 and 140 MeV. Note that the virtual correction 6hard i independent of this cutoff parameter. Note also that the results depend on the choice of electron scattering angle as well a.,3 on Q, and that NE18 kinematics were used for these calculations. The final colum n is the percentage contribution of the proton-proton and electron-proton interference terms to the total Bremsstrahlung correction ( 6,Pft+6,PPfI ). 650ft Q2 AE bh.d bee soft bep soft 6PP soft bsoft (GeV/C)2 MeV I 10 140 -0.07 0.332 0.158 0.015 0.007 0.007 0.003 0.354 0.169 6.2 5.9 3 10 140 -0.08 0.377 0.190 0.038 0.020 0.019 0.009 0.434 0.219 13.1 13.2 5 10 140 -0.08 0.398 0.205 0.056 0.030 0.028 0.014 0.482 0.249 17.4 17.7 7 10 140 -0.09 0.424 0.226 0.070 0.038 0.035 0.019 0.529 0.283 19.8 20.1 136 % ep+pp Psoft Table 49: Single photon Bremsstrahlung spectrum, evaluated at two NE18 kinematic settings and integrated up to photon energies of 10 and 140 MeV. bs.ft is calculated using the full SPA expressions of Equation 428; ' is from the closed form expressions of Equation 442 found in the ultra-relativistic limit. The final column present the percentage discrepancy of the UR calculation relative to the full SPA. Q2 AE oft 6' discre 6S soft (GeV/C)2 MeV I 10 140 0.354 0.169 0.347 0.166 -2.0 -1.8 10 0.413 0.474 -1.7 140 0.249 0.246 -1.2 5 % P. turn transfer and intial final fermion momenta are large compared to both the nucleon and electron mass. In this limit, one obtains the following closed forms: bur ee Ur 7 Z2 PP bur ep kk I (AE)2 a -In a 7r z _' 7r In P0P 01 In (AE)2 0 Of In P P AE2 -q 2 M2 In k In - k' q2 I M2 + In kki AE2 -In 2 2 P Of M k In - k' I kk' 2 M2 +-In k In - k' .(4.42) These forms reveal the essential features of one-photon emission: all of the dependence of bsoft(AE) on AE takes the form In AE but additional terms independent of the photon energy cutoff are also present. These expressions will prove very useful later on, and so it is worthwhile to see how good the UR approximation is at NE18 kinematics A comparison of 6,.ft(AE) computed using Equation 428 and Equation 442 is presented in table 49. One sees that at NE18 kinematics, the approximation is accurate to at least 2%. 137 4.5.3 Higher Order Bremsstrahlung In the previous section, we removed the infrared divergence from the first order Brernsstrahlung cross-section. The other divergence that needs to be understood occurs in the limit AE --+ 0. This indicates that there is an infinite possibility for emitting a photon of infinitesimal energy. In other words, the first order perturbation expansion breaks down as AE becomes very small, and one must include the possibility to emit many soft photons. In actuality, the probability of scattering without losing any energy to Bremsstrahlung is zero so the actual cross-section should approach zero as AE -+ 0. It was originally determined by Yennie, Frautschi, and Suura (ref 6 that the emission of soft photons can be summed to all orders via exponentiation: do, (WO< AE = duM dQ, ' dQe e- 6,.ft (A E) ( - (4.43) bh.d) ep The notation (2) indicates that this expression represents the cross-section for emitting any number of soft photons, each with energy less than AE. In practice, however, one is interested in the total photon energy emitted. This case is discussed below, and found to agree with the preceeding formula to within a correction of order a 2 Recall that the probability for emitting a single Brernsstrahlung photon has a energy dependence that factors from the angular distribution A(L) (Equation 422) order to maintain a handle on the AE - In divergence for the moment, we write the cross- section to emit one photon with energy wo > AE, along with any number of photons each with energy less than AE: do, (n = , AE) dQdL,),OdQ, Here, du dQ, ,P e- &.ft(AE) (I - h.d) 0 0(w - AE) (4.44) W 1 represents the usual step function, and dQ1 indicates the emission angle of the 138 photon wl. Similarly, the cross-section to emit two photons with energy Wo 1 > AE and W2 > AE, along with any number of photons with individual energy less than AE is do, _(n = 21,AE = do, dQdw0dQjdLo0dQ2 1 2 dQe (4.45) I - 6h.d) P I A Lj )0(w 2 Lo0 1 - AE) LIJ0 2 0(w - AE). 2 Generalizing this to the case of n photons of "large" energy, one obtains: do, ffiedwoffij (n, AE = do, ... dwOdQ, n dQe (4.46) e- 6..ft(AE)( I - 6h.d) ep A j n! O(Lv - AE) ... u 0 I Wo n O(wo - AE). n At this point, we can impose a requirement Ejw9t < E on the total rather than individual emitted energies by replacing the step functions above with a delta function and integrating over individual photon energies and angles. Also, we sum over all numbers n of emitted photons: do, JAE) Ef dQdE n=O E AE 0 dw dQj- E AE dwodQn n da :Td-jj (n, A E) dQdw'dQj 1 n n 6(w1 + - + wo n - E). (4.47) One observes that the angular integration can be done at once for each photon, and for convenience we write A dQ, A ). (4.48) We then combine equations 446 and 447 to obtain do, (AE = do, dQdE dQ, n e_6..ft(AE) ( - h.d) c'o ep n=O n i=1 E & A 6(w + - + w - E). E UP n (4.49) 139 This is a form that we will encounter again later on. It can be evaluated by substituting an integral form for the delta function: 6(E wo i - E = I 00 2z eiX(F WO-E )dx, (4.50) i and observing that Equation 449 has the familiar form n=O ez. Carrying through the computation, one finds that the AE -- 0 divergence in e&-ft(AE) is cancelled by the similar terms due to the AE lower integration limit. Taking the limit AE --+ 0, the following relatively simple form is obtained: du (AE = do, dQdE -b'.ft(E)) e6.ft(E)F(A). (I - 6h.d) dQ, (4.51) S The function F(A) is expressed in terms of the gamma function and Eulers constant C '_ 0577; if we recall that A is of order a, we can expand this function in powers of A: F(A) r( e-c'\ A) r2A 2 12 + (4.52) Integrating Equation 451 over dE, one thus gets du (1: dQ LO0< E = do, lep( - bh.d)e- &.ft(E)[+ O(Cf2)], (4.53) dQ, which agrees with the previous exponentiated formula, Equation 443, to within a correction of order a 2 Exponentiating &ft thus provides a good approximation to the Bremsstrahlung crosssection for emitting a total photon energy up to a certain cutoff value. The exponentiated cross-section also has the correct limiting behaviour, liME-0 dQ, d (E 140 Lo' < E = 0, since bsoft(E - B(pipjE) - ln(E). Note, however, that 6hwd i not exponentiated. Mo and Tsai [58] take the point of view that whether or not to exponentiate this term is an open question. As with the choice of which second order diagrams to include in bhd, the crux of the matter is that experiments comparing results with one another must use the same prescription. In the case Of the correction. itself is small: ( and - 6had) 4.5.4 c6h.,d 6hwd, however, this is generally a moot point since varies from 007 to 009, and so the difference between bh,,,d is at most 04%. Peaking Approximations We have now calculated the energy distribution for multi-photon Bremsstrahlung to all orders, given the soft photon approximation and to within an order a 2 correction. However, to alculate radiative effects in a coincidence framework, one must know the effect of the emission cross-section on all measured particle vectors. The integrated probability up to an energy cutoff is not enough, and one needs to know the angular distribution f photons as well. The angular distribution of single photon Bremsstrahlung is given by Equation 423, and is plotted in Figure 44 for Q = 7 and 15 (GeV/c)'. One salient feature of the distribution is immediately apparent: the radiation is strongly peaked along the directions of the incoming and outgoing electron. Only a very broad peak is seen in the direction of the scattered proton at Q = I (GeV/C)2, but it becomes more sharply defined as Q2 increases. These features suggest a simple approach to the angular distribution, known as the "peaking approximation": the single photon Bremsstrahlung spectrum may be divided into three discrete photon directions, along each of the vectors k, k , and p' In other words, we replace A(c) in Equation 422 with the simple form Apeaking(L = A,6(C - k + A,,6(c,- k + Apb(L- '/), 141 (4.54) where f dQ,b(c) The terms of the one photon angular distribution A(c) may be divided into three groups, due to the electrons, the electron-proton interference, and the protons respectively: kf 2 A ) awo, w 47r2 ki kf -2 2 W ki kf ki w kf w k- + Pf W. Pi f LO Pi 2 Pf PI W - Pf W Pi (4.55) In order to better understand the structure of the peaks, consider the expansion of the first term in an angle describing the direction of photon emission relative to the direction. In the region < k Jkl, 2 kf k, 2 LO 4- '." w kf w ki Iki 14o2, (4.56) M4 indicating that extremely close to the k peak), the emission probability actually drops to zero. This feature is too small to be seen in the electron peaks of Figure 44, but is apparent in the much broader proton peak at Q2 = 7 (GeV/ C)2 (since Further away from the peak, in the region - IWI < VI is of order 1). the angular distribution falls off quadratically with : kf LO2 We will refer to this ki kf L, - ki 2 4 02' (4.57) shape later on. We next need to determine the values of A,, A,,, and A, by integrating the various terms of A(L) and distributing the results among the three peaks. The first (electron) 142 loI 10 10i lo 10 i01 101 10 10I 10 lo I 10I 10 10 10 10 10 10 10 10 10 10 10 11 io9 108 107 1(6 105 104 103 102 10I T0 10 -1 lo -2 10-3 10. -14 photon angle Figure 44: Angular distribution of first order Bremsstrahlung photons from Equation 4.23, calculated at Q = 7 and 15 (GeV/c)' and showing the improvement in the peaking approximation with increasing momentum transfer. The photon angle is measured with respect to the direction of the incoming electron and given in degrees. The directions of the scattered electron and proton are indicated by dotted lines and the notation , O 143 term of Equation 455 produces two terms of the form k2 Cew02 j( dQy (W a k)2 - - 7r (4-58) (one for each of k and k). Since the integrand is highly peaked in the direction k (or W), it is assumed that all this strength contributes in the k (or W) direction. Next consider the integral of the cross term, aw 02 2 k V 47r2 IdQ1f (4.59) (w k)(u., k)' In this case the integrand peaks in the k and k' directions. We evaluate it using I &21 (w k k' k)(w k) 1`1 k - k' = I dQ1fP k)(k k wO = 2r In LA)2 k - kl Jkl + 2r + I d, In LIJ2 k - k' (k' k1 + WI klo - k/I k)(w IWI k) wO (4.60) This expression approximately integrates over the two peaks separately; the first and second integrals are assumed to contribute to the k and k' peaks respectively. Cornbining these equations, one obtains the "typical" peaking approximation for electron Bremsstrahlung: A = a 4k2 - [In _ 11 r M2 Ael = - a r [In 40 M2 _ 11. (4-61) We can further assume that the third term of Equation 455, although only broadly peaked at intermediate energies, contributes entirely to the final proton peak, yielding Apt = a [In 7 ePO+ 0 - 10 144 - 21. (4-62) Some Bremsstrahlung strength still remains, due to the electron-proton interference term of Equation 455 and to the non-peaked contributions missed by the approximation of Equation 460. This is true even in the ultra-relativistic limit, where one expects the peaking approximation to be the most valid (cf. 44). If one uses the closed form UR limit expressions of Equation 442 to determine the difference t (El) - &ft(E2) between two energies, and compares this with the result using only the peaked strength described by Equation 4.61 and 462, one finds two missing terms. These are In 7r E2 4 In El due to the electron-proton inteference term a In E2 7 El due to the non-peaked strength in jkj (4.63) jk1j and 2 In 1 - Cos(0,) (4.64) 2 .Our approach is to preserve the total strength (as bee evaluated in the UR limit) by distributing the contributions of these non-peaked terms among the three photon peaks. We choose to split the two terms evenly between the electron peaks: A, a k - 2 In(-) 7 P = A, Ae = Ae Apt AP/. = + a 7r k 2 In(-) ki In In 1 - cos(O,) 2 I Cos (O,) 2 (4.65) This set of formulae can be termed the "extended peaking approximation" for single photon Brenisstrahlung. In the text below, the notation A will be assumed to mean . From Section 45.3, we know that including higher order Bremsstrahlung is critical in 145 evaluating the energy spectrum for low photon energies. One is then led to consider its effect on the angular distribution. Calculating such higher order contributions directly from Equation 423 is a formidable task. Instead, we observe that the single photon peaking approximation, do, dQdw k + Ae'6(C - k) + Ap'6(L do, doe epwO (4-66) effectively provides us with three independent single photon energy distributions, each for radiation in a fixed direction. We can then proceed in the manner of Section 45.3 and determine a multiphoton spectrum, this time in terms of three energies: the total photon energies E,, Ee,, and Ep, emitted in each of the three peaked directions. The total radiated three-vector is then simply Wtotal--:":Ej + Eek + EP. (4-67) Furthermore, radiation along the direction of a given particle can be interpreted as radiation due to that particle. In this way we correct the q vector used to evaluate d, dQ, ep at the scattering vertex for energy radiated before the scattering (i.e. radiated by the incoming electron). This was seen in Section 45.1 to be the source of the largest discrepancy between the soft photon approximation and full calculation for single photon radiation. By analogy with Equation 449, we obtain the cross-section to all orders for radiating a total energy Ee along k, E along W, and Ep, along p', as well as any number of soft photons with energy less than AE: do, do, dQdEdEedEp, (AE = dQ, ep e- 6soft EEL ':'o (I - h.rd) 'o 00 1=0 m=O n=O 146 A E b(We C dw"O E i Oi M E i n n! 1 A Leto aE + L',e _ Ee) + bpe, 0I +... + Le 10 Eet i Welo EP, A 1 dL,)P i=l 0 10 Io 6 (L"P I+ -- + )Pn L')P Using the same technique as in determining Equation 451, one obtains: da do, dQdEdE,,dEp, e- 6..ft (afl (I - dQ hwd) e AC1n(EC/AE) p • eAe11n(E,,/aE) • eAP1 1n(EP,/aE) '\PF(Ap,). A,'F(Ae,) Et Ep, Again, the As are of order a, and so F(Aj) (Equation 452) is of order a 2 ACF(Ae) E, (4-69) to within a correction 1 We see that the In(AE) dependence of b,.ft(AE) will be cancelled by the other terms of the expression, taking care of the AE -* 0 divergence of the single photon spectrum. By construction, the subdivision of the terms of s of the extended peaking approximation provide a ft(AE) which depend on AE: b.oft(El - boft(E2) = In E2 ( + + (4.70) El (Note that this is true only in the UR limit). However, 6,.ft contains additional terms. Using Equation 442, we find that these terms can also be subdivided in terms of the As: 6,(AE) A,In(. AE ) bel(AE) A,,ln( bp (A E) AptIn( VkO k'O 147 ) AE AE (4-71) Employing these definitions, we can take the limit AE 0 to produce our final result for the multi-photon peaking approximation: do, dQdEdE,,dEPl do, dQ, ep Xe _1,!,E,, do, dQ, (I 6P, EP,) be (Ee) P 6h.d) ep (4.72) Ae A,, AP, X Vkk I) A.!(/kk,)A_, (,,Me) AP'Ele +A"El+ el Al-/EplI+AP' The cross-section thus factorizes neatly into three independent functions, for the total energy emitted in each of the three radiative tails. The angular distribution implied by the above equation can be evaluated easily by a Monte Carlo program such as our PWIA calculation, by randomly generating the energies emitted in each direction and adjusting the fermion vectors accordingly. However, it is worth studying the multiphoton angular distribution analytically, to determine the approximate shape of the multiphoton peaks. For our calculation to be valid (or useful), we must confirm that these peaks are substantially broader than the single photon peaks, which were approximated as delta functions in Equation 454. To accomplish this, we employ a change of variables: from E,, E,,, and Ep, to E, u, and v Here, E is the total radiated energy E, + E, + Ep,, while the emission direction is fixed by u and v: U V Eg Ei Eg EP, Note that u and v vary from to oo with u, v (4.73) oc corresponding to emission in the e direction, u --+ 0 corresponding to emission in the e' direction, and v - 0 corresponding to emission in the p direction. The Jacobian between these two sets of variables is 148 straightforward: dEdE,,dEp, - dEdudv EE,,Ep, (4.74) Euv Consequently, the multiphoton emission cross-section 472 can be rewritten easily in terms of the new variables. The dependence on the total energy E factorizes completely from the angular distribution, and the integration over emission angles can be accomplished, yielding do, - dQedE d API - - ' (I - d,,p h.d) (VMP v/kk I)A Vkk 1)), x x (A, A,, Ap,)r(I + A,)r(I + A,,)r( r( + A, A,, + Ap,) 4.75) Apt 9 Recalling that the As are of order a, one finds that the ratio of gamma functions on the last line is I to within the usual O(a') correction. To within this accuarcy, this distribution agrees with the previous multiphoton formula, Equation 451. The analysis of the photon distribution simplifies greatly if one neglects proton radiation. Taking to be the angle between the photon and k, and 0,, to be the usual scattering angle between k and W, one finds that for do, < 1, sin(O,,)-'--' dQdEdO (4.76) 01-A'/ and for 0 -- 0, < 1, do, dQdEdO sin(O,,)-", ( _ The photon spectrum thus drops away from the peaks at the rate more gradual than the (4.77) ")1-A,, As this is falloff of the single photon peaks, our calculation of the multiphoton distribution from perfectly peaked single photons is reasonable. 149 In the case of proton radiation, of course, the peaking approximation is suspect from the very beginning. Its use hinges on the relatively small Brernsstrahlung contribution of the proton, and on the resolution of the experiment. Also, as pointed out at the beginning of this section, at sufficiently high photon energies all radiative tails converge on the same k = , kinematic path. The sensitivity to the precise angular distribution is thus most apparent at low photon energies. The effect of the peaking approximation on the NE18 PWIA calculation will be examined in Section 45.6. 4.5.5 External Bremsstrahlung One element of the radiative calculation remains to be described. This is the spectrum for the emission of Brernsstrahlung photons in the field of nuclei other than the one participating in the hard scattering. These losses occur as a charged particle moves through the target material and traverses vacuum chamber windows and air gaps. The probability distribution a particle of momentum k to radiate at total energy of E"' when traversing t radiation lengths of material is found to be 33] 1 r(i+ bt E"t bt) Eext Jkl Eext bt text Jkl (4-78) where the parameter b depends on the atomic charge Z of the target material: b = LI = L2 1 9 12 + ZLI + L2 I n(184.15 - - n(Z) 3 2 = ln(1194. - - n(Z). 3 150 (4.79) The function V` is a correction for large photon energies, expanded to second order in E ext IkI ext (X) X + 3 X2 (4.80) 4 External radiation is far simpler to treat than internal. First of all, the particles radiate independently and so incoherently, and this eliminates the non-peaked strength cause by the interference terms of internal Bremsstrahlung. Furthermore, proton radiation is suppressed relative to electron radiation by the factor (_)2 M and so can 10-6 , be neglected ntirely. Equation 472 can thus be expanded in a straightforward way to include the contributions from external radiation along the k and k' directions: do, dQdE"1tdE'PxtdEintdE"t = i f f do, doe ep(I - h.,-d) Eext X r(I + bI X I X btf F(I + btf) EXt f bti A- nt E, X Sent Iki Efext IkI k btf Af I E int X - f Eint f Nkki Af (4.81) Here, the internal proton contribution has been omitted for convenience, and the subscripts i and f have been introduced to indicate the initial and final electron arms. Since both it and Ext are emitted in the same direction, we would like to rewrite te distri- bution in terins of the total energies Ei and Ef radiated along k and W. This problem is exactly analogous to the transformation made between Equations 472 and 475, where a change of variables was made from three energies E,, E,,, Ep, to a total energy E and angle variables u and v. The result is do, dQdE,nYExtdEintdEext I f f - do, (I doe eP 151 6hard) F(I + bti) (I + btf) (bt + Ai) X kbt (vl-kkl),\i (btf + Af) dEj k1btf (Vlrkk1)Af Eil-Ai-btj dEf r'I-Af Ef .(4-82) -btf We thus see that the As of internal radiation play much the same role as the material thickness bt of external Brernsstrahlung. One can also express the external radiation contribution in terms of the usual Bremsstrahlung functions 6,,ft. One obtains forms which are very similar to those of Equation 471: 6ext (AE e = bex el t(AE bti In( AE) k k' = btf In( AE). (4-83) These functions can simply be added to the corresponding b(AE)s for internal radiation in Equation 472, yielding the same result as Equation 482. Thus far, the correction function the ratio E-1 jkj Oext e-t) has been neglected. At NE18 kinematics, Jkl in which the function is expanded is small only the first order term of Oex, (x) 0.1). Consequently we take and include it in Equation 481. Carrying through the angular integration, one obtains multiplicative factors Vt and Vxt f to include in Equation 482: Dext (Ei) - i (The same form applies for 4.5.6 (Dext with f bti bt i Ej A t (4.84) ki f everywhere). Radiative Techniques Employed 'in the PWIA Calcula- tion Two models of the radiative corrections are included in the PWIA Monte Carlo SIMULATE. The first uses the multiphoton energy distribution of Equation 453, evaluated using the full SPA expressions of Equations 437 and 438. The angular distribution is 152 taken to be the pure peaking approximation of Equation 454. The strength is distributed among the three tails Z= subscripts e, 23 using the fractions A' (I is shorthand for the usual tail I '). The second method tries instead to obtain the correct multiphoton angular distribution by generating the total photon energies E, E, Ep, emitted along each direction., and summing the resulting photon vectors according to Equation 4.67. The distributions are generated according to the independent forms found for each tail in Equation 472. These energy distributions were calculated using the approximate closed form expressions of Equation 442, found in the ultra-relativistic limit of high momentum transfer and particle momenta. These two choices represent a trade-off between the bt:st available forms for the photon energy (first technique) and angular (second technique) distributions. The first method can thus be referred to as the "peaking" technique, and the second as the "multiphoton" technique. Note that these names are somewhat misleading: the "peaking" formalism certainly involves contributions from Bremsstrahlung radiation to all orders, and the "multiphoton" prescription involves the peaking approximation at the one photon level. One hopes, of course, that the two prescriptions give very similar results and this indeed turns out to be the case. The distributions of counts calculated by- SIMULATE using the two techniques are sufficiently similar that one is hard pressed to see any differences on a plot of the projections in E, and p,. The integrated yields are less than 1% different at all Q. This excellent agreement indicates the lack of sensitivity of our results to the fine details of the photon angular distribution, and the validity of the UR limit at our energies. Comparisons of the radiative calculation with the data will be presented in the following chapter. 153 Chapter Results of the Experiment 5.1 Extraction of Results 5.1.1 Acceptance Cuts The symbol is used throughout the following discussion to denote the experimental ac- ceptance. This acceptance was defined by cuts placed on both the reconstructed particle vectors and on the extracted values of E The p and p,. coverage is, of course, different for each setting of Q and ',P and to a lesser extent for each target. Cuts in p, were chosen at each of these settings, to exclude low statistics bins at the edge of the phase space. Overall kinematics-independent E, and p,,, cuts were also imposed on the results and are listed in Table 5.1. For the carbon, iron, and gold targets, p, values on only the positive side of the q vector are selected because the negative side was not fully sampled at all Q' settings. The 250 MeV/c choice of upper limit was based on the maximum coverage available at all Q, without including regions where the phase space did not extend to the Fermi energy in E,. This cutoff is reduced to 210 MeV/c for the gold data, because of the slightly smaller phasespace available for the Q2 of 7 (GeV/c)' data. Note that alternative choices of overall cuts 154 Table 5.1: E, and p, cuts placed on results at all kinematics. Target 111 2H 12C 56Fe 197 Au E, cuts MeV p,,, Cuts Mev/c -25 -25 -25 -25 100 100 100 100 -160 -160 0 0 160 160 250 250 -25 100 0 210 were tested as well and used to provide uncertainties on the results. These are discussed in a later section. The cuts on reconstructed spectrometer quantites are designed to keep these vectors inside the well-understood regions of the spectrometer acceptances. Such cuts are particularly important in the momentum variable , as the acceptance is not flat outside ± % in either spectrometer. Perhaps more importantly, the optical matrix elements were only tuned in this region and are known to produce unreliable results at larger momentum deviations. Also, evaluating our results using several cut settings supplies an estimate of the systematic uncertainties caused by our spectrometer models. Four cut settings were used in the analysis, and are listed in Table 52; CUT 3 was used for the extraction of our final results. 5.1.2 Extraction of the Spectral Function The spectral function was extracted from both the data and the results of the SIMULATE Monte Carlo. Both sources produced histograms of experimental counts, binned in E, and p,. Such a distribution is related to the differential reaction cross-section by the relation N(Em7pm) )CKeff I6(E,,,,,-) 155 d6o, d" V, d6V (5.1) Table 52: Cuts placed on reconstructed spectrometer quantities. Note that the same cuts were imposed on both the 16 GeV/c and GeV/c. refers to the in-plane angle (relative to the magnetic axis), and D to the out-of-plane angle. CUT 3 was used for all final results. CUT (D % where N(E, mrad mrad I 2 ± ± ± 25 ± 60 ± 25 ± 40 3 ± ± 15 ± 40 4 ± 4 ± 15 ± 40 PM) is the array of true coincidences corrected for experimental efficien- cies, and dV denotes the six-fold differential phase space of a coicidence experiment dc'dQ,,dE'dQP1. V(E,,,, p,) indicates the total phase space volume contributing to a given (E,,,,p,) bin. Using the PWIA factorization assumption of Equation 17, one can remove the spectral function S(E,,p,) from the integral and obtain N (E,,,, PM = L S (E,, PM)K oep I Empm) P (EM I PM) Here, Ko,,p (Empm) (5.2) is the average off-shell ep cross-section, averaged over the events collected in an (Em, PM) bin. Taking this average and removing KO',p from the integral is reasonable because this function is slowly varying over the quasielastic peak. 'P(EM, p"') is the so-called coincidence phase space, given by P (Em, PM = JV6 d 6V, (E,.,,,Pm) (5-3) and represents the probability density for obtaining a given Em and PMover the six-fold phase space of the experiment. It is evaluated by running the Monte Carlo program SMULATE without radiative effects, and weighting each event with Wg,, only (see Section 156 4. 1). The physics is thus removed from the calculation, leaving only an integral over the phase space. The P(E,,,, p,) array is finally multiplied by the 6-fold phase space volume in which the generation is performed, and divided by the total number of generated events. The extracted spectral function is thus defined as S (EM PM) = N (Em, M) (5.4) IcKo,,p (E,,p,,,) P(EmpM)' 5.1.3 Extraction of Transparency The kinematics of the experiment involve data taking at several setting of the proton spectrometer angle for each Q. The transparency at each of these settings (denoted ) is defined to e T = E R (E,,,,p,,,)Ndata(Em, M) (5.5) E Ri(E,,,,p,,,) NPW IA Em PM)' The values at individual angle settings were then combined to obtain the most probable result for the net transparency at a given Q: I T- T = Ei I:i dY.7 -1 . (5.6) dT.7 The errors dTj are the sum in quadrature of the statistical and systematic errors. The fractional statistical error is taken to be the familiar VN standard deviation of a Poisson distribution. The systematic uncertainties are discussed below. 5.1.4 Systematic Uncertainties A summary of the systematic uncertainties of the experiment and their effect on the extracted transparency is given in Table 53. Most of the uncertainties have been presented and explained in previous chapters, but some appear here for the first time and 157 require additional comment. The error in random subtraction was evaluated by looking for systematic trends in the number of coincidences found between E, of 25 and -5 MeV. There should be no spectral function strength in this region. Events found in this region should be taken into account by statistical errors; the uncertainty listed here is an upper value on any systematic deviation from zero. The acceptance error was determined by examining the fluctuation in results for different spectrometer cuts. The error on radiative corrections was computed from two sources. First is the variation of the transparency with the upper E, limit used in the sums of Equation 55. Upper limits of 50, 80, 100, and 130 MeV were applied to the hydrogen and deuterium data, and produced variations of < 2 %. This is referred to in the table as the error on "internal radiation", as radiative corrections for these thin liquid targets is dominated by the internal portion. A similar test done with carbon and E, cutoffs of 80, 100, and 130 MeV confirmed this result, although the Lorentzian tail of the carbon missing energy distribution also contributes at high E, preventing the isolation of purely radiative effects. The "external radiation" error was determined using test runs performed at iron targets of different radiation length 2% and 6 Q = I (GeV/c)' with carbon and r.l. for carbon, and 6 and 12% r.l. for iron). Assuming that the target thicknesses are well known and the experimental luminosity properly computed, external radiation is the only element of the analysis affected. An error of 2 was determined from these comparisons. The uncertainties due to the model spectral function and o,p were simply determined by performing the PWIA calculation with various formulations of these. The choices of prescription for the ep cross-section and model spectral function are explained in Sections 42 and 43 respectively. The individual systematic uncertainties are summed in quadrature at the bottom of the table. The net uncertainties for 'H, 'H, and `C are 78% in magnitude, and are 158 dominated by the experimental acceptance errors. For the heavy targets 56 Fe and 197 Au, however, the systematics are considerably larger (I 1 13%). The relatively poor knowledge of the model spectral functions for these targets is responsible, as well as the sensitivity of the transparency to modifications of these input functions. The transparency depends on the number of experimental counts determined by the PWIA Monte Carlo, and this in turn is the result of folding the spectral function with the coincidence phase space. The IPSM spectral function is normalized to Z S(EM Pm)dEd S(Em, pm)dEmp2 M dpmdQpm, 3P, (5.7) and so produces a constant result when folded with p2M . The coincidence phase space, however, has a different p dependence in general. Consequently, adjusting shell model parameters which affect for example the width of the spectral function may cause large changes in the number of counts computed by the Monte Carlo of the experiment. 5.2 Hydrogen Results The spectral function is not a meaningful quantity for elastic scattering from free protons, as the differential cross-section is inherently two-dimensional. However, the distribution of hydrogen ata counts in Em and pm compared with the Monte Carlo calculation provides a precise test of many aspects of the calculation. The true distribution of elastic events is precisely localized at E = and p = ; any deviation from this must be due to the improper subtraction of randoms, experimental resolution, and particle radiation. The last two effects should be correctly modelled by the Monte Carlo. In particular, a comparison of the data and Monte Carlo on the hydrogen radiative tail provides a precise test f the radiative procedure, unclouded by other physics. Finally the hydro- 159 Table 53: ment. Summary of systematic uncertainties affecting the transparency measureItem dead times % Uncertainty in T 1.4 deff 1 proton absorption target thickness liquid 2 solid charge tracking, 16 GeV/c tracking, 8 GeV/c multiples, 16 GeV/c multiples, 8 GeV/c random subtraction acceptance internal radiation external radiation O"P 0.7 0.2 0.5 1 1 1 1 1 5 2 2 2 model spectral function 12C 2 6 9 12Fe 12 Au correlation corr. 12C 3 6 6 12Fe 12 Au sum in quadrature - 1H 2H - 12C - 12Fe - 7.0 7.0 7.9 11.0 12.9 12 Au 160 gen transparency must be one and this provides the ultimate test of the experimental normalization,. The distribution of hydrogen data counts as a function of E, is presented for all Q2 in Figure 5-1. Superimposed on these figures is the corresponding Monte Carlo calculation. Also included here are the E, distributions for deuterium (Figure 52), as the single deuterium bound state is very sharply peaked at the binding energy of 22 MeV and so 'H data in this coordinate provide the same precise test of the radiative procedure as 'H. The figures clearly demonstrate that the radiative prescription perfectly describes the data to within its statistical uncertainty (note that the statistics available for 'H and2 H are higher than those for any other nucleus). The hydrogen transparency is listed in Table 54 depicted in Figure 513 and is consistent with unity at all momentum transfers. As a quantitative measure of the E and p, dependent agreement discussed in the previous paragraph, one can evaluate the hydrogen transparency with a variety of E cuts. One finds that the transparency varies by an amount well within the statistical error of the data for upper E cutoffs from 50 to 130 MeV. The statistics provide a precision from 1% at Q = I GeV/C)2 to 4 at Q2 = 7 (GeV/ C)2. 5.3 Spectral ]Function Measurement The spectral functions extracted from our data are represented here in their projected forms, p(p,) - f SdE, andC(E,,) =_f Sdp,, in figures 53 through 5-10. The PWIA calculation is normalized to the measured transparency. In all cases, only statistical errors are shown. These can be inflated by the amounts indicated in Section 51.4. First, we see that there is reasonable agreement between the measured E distribu- tion and the PWIA calculation. In particular, the carbon spectrum shows no evidence of 161 103 104 V) 1 C 10 :3 0 0 -2 C 2 :3 0 10 0 100 10- I E,,, (MeV) E,, (MeV) 10' 102 101 U) 102 0 S 0 0 S 0 U 100 10-1 0 50 100 0 E .. (MeV) 50 100 Ern (MeV) Figure 5-1: Distribution in E, of coincidence events recorded for the hydrogen target, compared with the prediction of the Monte Carlo program SIMULATE 162 I lo" 10 3 102 10 2 U) -!2 C 73 0 U 1 :3C 10 0 0 10 0 100 lo- I E .. (MeV) E M (MeV) 103 10 2 102 rn -3C: U) -.1 -3C10 I 0 0 0 U101 100 100 0 50 100 0 Er,, (MeV) 50 E rn 100 (MeV) Figure 52: Distribution in E, of coincidenceevents recorded for the deuterium target, compared with the prediction of the Monte Carlo program SIMULATE. 163 .08 .08 .06 .06 I I ID :2 a) M 604 04 a11 In0- V) (A '--.02 "-.02 .00 .00 a a Em (MeV) EM (MeV) .08 .08 I . I . .06 I I . Q2 -1 I I = 68 I I (GeV/C)2 .06 I I Q) :2 20 604 - eO4 a- CL 0 M V) (n '--.02 '--.02 .00 .- - a nn . . . . . .. . .- .- . .- .0 E.. (MeV) 50 .- -I .---- -- 100 E M (MeV) Figure 53: The extracted spectral function for 2H, integrated over dp,,,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 164 .06 .06 .05 .05 04 -,.04 0 M 03 9 Q) M 03 C C1. -11 0 (n 02 v).02 CL n 'a .0 1 .01 .00 .00 E. (MeV) E,,, (MeV) InC Ij: .06 - .05 .05 -': . . . I I I . . Q2 I I I I I I . = 6 . . (GeV/C)2 - 04 04 :Z 0 M 4) :2 503 03 E 011 na- (n.02 v).02 11-1 a a I-N .01 -: .01: I .00 .00 E.. (MeV) . - .-II 0 I- -50 I E. - II 100 (MeV) Figure 54: The extracted spectral function for "C, integrated over d3p,,,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 165 .10 .10 .08 .08 I I Q).06 Q).06 E I0.04 a-E aI0.04 V) V) .02 .02 .00 .00 E,,, (MeV) .10 .10 .08 .08 I I 0.06 0).06 E. (MeV) E (MeV) aE E 0- I0.04 I0.04 V) V) 1-1 .02 .02 .00 .00 50 0 E -, 100 50 0 (MeV) I - I M 100 Figure 5-5: The extracted spectral function for "Fe, integrated over dp,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 166 .12 .12 .10 .10 I .08 1 .08 Q) M 4) :2 - E06 506 0011 CL 110 v).04 v).04 .02 .02 .00 .00 0 50 100 E,,,, (MeV) E M (MeV) .12 .12 .10 .10 I .08 1 .08 Q) :2 0 :2 - E06 506 aa am in.04 (n.04 I--, .02 ,02 .00 .00 0 50 100 0 E.. (MeV) 50 E M 100 (MeV) Figure 56: The extracted spectral function for "'Au, integrated over d'p,,,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 167 11-1 lu I... I.-I .... I''.. : Q2 10 -6 10 : 1-1 -7 >01 0 :2 - E E LU W 0 '10 0 -a Inio : r 10 -9 . ................- . I -2 DO p,,, (MeV/c) -0-I ...... I......... Q2 10. = 3 GeV/c) -6 I 2 10 - 0 p,,, (MeV/c) 100 2( 10 -0I .................. = 68 (GeV/C)2 10. -6 _ r- I U '-I 110 E ,,lo . I I I . I I I . . . I . . . I -100 Q2 I1 0 1-1 .-7 110 W 'D (GeV/ C)2 I U U I--, 110 '!e 10 = .-7 E Li 0 -8 '-8 1110 1-1 k-- -9 -9 10 10 -200 F-- -100 ............. 0 p"' (MeV/c) 100 I 200 -200 -100 p, 1 .6 . . . .ido' ' ' 200 (MeV/c) Figure 57: The extracted spectral function for 2H, integrated over dE,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 168 p .. (MeV/c) p"' (MeV/c) 10-7 II "I I U 'I-, 111 Q) -8 4) -a 7 0 -:M0 E L'i E W a 0 V) (n I--, I--, 10-9 10-9 r) .. (MeV/c) p.. (MeV/c) Figure 5-8: The extracted spectral function for "C, integrated over dE,,,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 169 10 n 40 U '_1 Q) E LLJ a V) I___% 10 P,,, (MeV/c) p .. (MeV/c) lo-' 11 -Io '_1 -7 4) t E LLJ a cn I--, 10-9 -200 100 0 100 200 300 p,,, (MeV/c) p,,, (MeV/c) Figure 5-9: The extracted spectral function for 'Fe , integrated over dE,,,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 170 10 I1 lu 1-1 :t :,Do E W a cn 10 pi", (MeV/c) p,,, (MeV/c) 10 10 0 Q 11-1 4) - ZO 4)- ZO E E W Uj a a (n (n 11 10 11-N 10 P,,, (MeV/c) p. (MeV/c) Figure 5-10: The extracted spectral function for "'Au, integrated over dE,. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 171 strength beyond that predicted by the PWIA calculation. This is in contrast to some previous measurements 10][12][14]made at Q of about 02 (GeV/c)' which found evidence for extra strength at high E,; this was interpreted as evidence of scattering from multinucleon currents which would tend to shift portions of the IPSM spectral function to higher missing energy (see Sections 13.2 13.3, and 54.1 for further discussion). As described in Section 13.3, it is possible that scattering from multinucleon currents is suppressed at Q2 above 1 (GeV/C)2 - a high energy probe is less likely to scatter from a compos- ite system that one of larger wavelength. However, we have seen how Bremsstrahlung radiation also shifts strength toward higher missing energies, and it is also possible that part of the high E, strength is due to radiation not completely accounted for. It would be interesting to use the Monte Carlo calculation and attendant radiative prescriptions developed for NE18 to reanalyze the older data. One also notes that there are some substantial discrepancies between the PWIA calculation and the data at low missing energy. This is the region dominated by the nuclear shells. The Is shell in carbon, for example, seems to be a good deal broader than that used in the model spectral function. The data distributions for deuterium at I Q = (GeV/ C)2 and for iron also seem broader than the Monte Carlo result. It is possible that the energy resolution of the experiment was underestimated by the Monte Carlo, a hypothesis supported by the discrepancy seen in some of the spectra at E below the Fermi energy (i.e. in the region where there is no spectral function strength and all contributions must be due to experimental resolution). Fortunately, the transparency measurement is largely insensitive to the precise shape of the E distribution, as the coincidence phase space of the experiment varies only slowly with missing energy. (This insensitivity was quantified in Section 51.4 above). The missing momentum spectra show good agreement with the PWIA calculation for 172 all nuclei and momentum transfers. Again, there is no evidence of contributions from processes other than those described by the PWIA. This observation also validates our use of a single absorption factor (the measured transparency) to describe final state interactions: Discrepancies between the calculated and extracted p(p"') would suggest the presence of significant elastic rescattering of the knockout proton, specifically rescattering which is sufficiently forward peaked to shift the proton momentum within our spectrometer acceptance. Such effects would have to be unfolded from the distributions using DWIA corrections (as is done in precision coincidence measurements at lower energies). Figures 5-11 and 512 show the spectral functions integrated over the approximate E, ranges of the 1p and ls shells. One sees that the good agreement with the Monte Carlo is maintained even over these restricted integration regions (i.e. there is no E,-p, correlated source of disagreement which might be hidden by the integrals over the full acceptance). The familiar shell structure is clearly in evidence up to Q of 7 (GeV/C)2 pip(p,) drops to zero at p = , while the s-shell strength peaks at p = . We emphasize at this point that the extracted spectral functions are not radiatively corrected. Rather, radiative effects are taken into account by the PWIA calculation, and the extracted distributions are always compared with the Monte Carlo results. The contributions to the spectral function from the various settings of the proton angle have been statistically averaged on a bin-by-bin basis, but one should note that the radiative corrections for a given E, p, bin are different for different kinematic settings. The spectral functions cannot be analyzed in a model independent way unless they are radiatively unfolded. Tis is the next step we plan to take in our analysis. The ability to radiatively unfold the data is a direct consequence of the radiative Monte Carlo, for SIMULATE can accurately determine the radiative tail shape in the measured quantities cor- responding to spectral function strength in any actual E, p,,, bin. This Bremsstrahlung 173 7 IL) 'I-, :Z :)O W E V) 10 p. (MeV/c) p,,, (MeV/c) p,,, (MeV/c) p.. (MeV/c) 10 7 0 "I. 0 ZO - E W 0 (n 10 Figure 5-11: Extracted p(p,,,) for the 1p shell of "C. The spectral function has been integrated over -1 < E,, < 25 MeV. The solid line represents the result of the PWIA calculation, normalized to the measured transparency; the data points are shown with statistical errors only. 174 10 10 M I 0 11-1 U 4) Q) :Jo :10 E W E LU V) V) a 'a ,--I 1C PI,,, (MeV/c) p,,, (MeV/c) 10-7 10 I 0 "I I "I 1-1 0 4) - :N -8 -:: 0 E E Ui Li I--, 1-1 'a (n 0 (n 10-9 10 P"' (MeV/c) p,,, (MeV/c) Figure 512: Extracted p(p,) for the Is shell of "C. The spectral function has been integrated over 30 < E < 0 MeV. The solid line represents the result of the PWIA calculation, ormalized to the measured transparency; the data points are shown with statistical errors only. 175 shape has all th e experimental acceptances and resolutions correctly folded in. The only model dependence in such a deradiating procedure would come from the radiation of spectral function strength into the acceptance from regions outside. A model spectral function is required to determine these contributions, but at the Q2 settings where miss- ing momenta on both sides of the q vector were detected, most of the spectral function is detected and these contributions should be small. 5.4 Nuclear ransparency Measurement The measured transparencies for all Q and all targets are presented in table 54 and plotted in Figure 513. uncertainties. The figure depicts both statistical (inner) and total (outer) Small differences between these numbers and those appearing in NE18 publications 62][63]are due to the averaging that was performed between these results and those of a parallel analysis performed at Caltech 40]. A detailed comparison of the transparency values with theoretical predictions follows, but some important conclusions can be drawn immediately. First, one sees that the expected signature of colour transparency - a rise of the results with increasing Q - is not observed within experimental errors. A gradual rise with Q2 is certainly permitted, but the results are consistent with a Q independent model of final state interactions. A notable exception to this, however, are the values at Q = I GeV/c)'. A dramatic rise in the transparency is exhibited at this momentum transfer for all A>2 nuclei. An immediate explanation of this effect is suggested by Figure 1-4, depicting the energydependence of the free pN cross-section. If one surmises that the pN interaction in the nuclear medium has a similar energy-dependence, one sees that final state absorption will be less for our Q = I measurement. At Q = 1, p is about 12 GeV/c and pp is significantly lower than the asymptotic 40 mb approximately achieved at our higher p' 176 Table 54: Summary of the NE18 transparency measurements. Statistical errors are given in parentheses after the total uncertainties. Target Q' (GeV/c)' 1 3 1H 0-99±.07 2H 0.91±.06 (.01) 0-65±.05 02) 0-51±.06 (.02) 0.42±.06 (.02) 12C -16Fe 197 Au (.01) 0-99±.07 5 (.01) 0.89±.07 03) 0.64±.06 03) 0.38±-05 (.02) 0-28±.05 03) 6.8 104±-08 03) 107±.09 (05) 0.91±.07 0-62±.06 0.41±.06 0.23±.04 03) 03) 04) 03) 0.93±-08 0-68±-07 0.44±-06 0.31±.07 (.04) (.04) (.04) (.06) settings. If one looks at 513 with an eye to identifying a rise with Q2, one is tempted to neglect the error bars and remark on the rise of the actual measurements. However, such an observation would conclude that a rise also exists in elastic scattering from hydrogen. It seems likely that a Q2 dependent systematic error in the normalization is involved, but is properly accounted for by the experimental uncertainties of Section 5.4. 5.4.1 Comparison with Glauber Calculations A search for an effect such as colour transparency must be made not only by comparing the measurements with CT predictions, but also with calculations made without including the effect. It was pointed out above that no evidence for a rise in T( Q2) is observed; one is then led to ask whether "conventional" calculations can explain the results. Such calculations are performed using the "Glauber" technique described in Section 1.3.4. Also demonstrated in that section was the variety of effects which can alter the calculated results and which must be taken into account. A large nmber of Glauber calculations have been carried out for a variety of nuclei, notably at A = 12 and - 200. A comprehensive selection of these is presented in Figures 5-14, 515 ad 516, along with the NE18 transparencies. For the sake of convenience, the following discussion will focus on the results and calculations 177 for "C. Before we 1.2 i 1.0 0. I--, N 0 I 0.6 - I-- 0.4 - 6OFe 197Au 0.2 0.0 0 I 1 I 2 I I 3 I 4 Q2 (GeV/c 5 Q2 for ease of illustration. 178 I 7 8 )2 Figure 513: The measured transparencies for all targets and points are offset in I 6 Q2 values. Some of the continue , however, we should mention the deuterium transparency measurement, which is consistent with a Q independent value of about 090. The applicability of a lauber- type multiple scattering calculation of final state absorption to the 'H nucleus, with its single spectator nucleon, is clearly questionable, and few estimates of the deuterium transparency exist. Two such estimates suggest values of 092 64] and 095 65], both of which are in good agreement with the data. The significance of our 2H transparency measurement remains an open question. Two immediate observations can be made from an examination of Figures 514 - 516. First is the large variation in the calculations. Much of this is due to the omission of different effects from the various calculations - in fact, it will be seen in the discussion below that no calculation includes all of the aspects of the problem that have been identified to ate. Second, one sees that the calculations are in general lower than the results. This suggests that the few theories producing high enough values of T( Q2 ) are identifying a important aspect of the FSI description. Only two of the calculations shown in Figure 514 are Q2 dependent. These are from Frankfurt, Strikman, and Zhalov (FSZ93) 701and Jennings and Miller 71] account for the energy dependence of the (free) pN cross-section over the NE18 kinematic range, while the rest use asymptotic values for this cross-section. The Q2 dependence of this effect must be understood if one is looking for a colour transparency signature. The asymptotic values of upArused also vary among the remaining calculations. The largest difference here is between the use of the total or the reaction (i.e. inelastic) cross-section. Most calculations use Ototal (with values varying from 40 - 43 mb). However, most calculations effectively work with a semi-inclusive transparency, integrating over all proton final states, and as is highlighted in the work of Kohama et al. (KYS93) 66], elastically rescattered protons will be recovered by such an integration. The reaction cross-section is less than 179 1.0 0.8 0.6 - - - - - - - - - - - - - - - - - - - - - - - - - - N 0 I_' I--- - - - - - - - - - - - - - - 0.4 - - - - - - - - - KYS93 a, - - - - - - - - - Nik93 KYS93 at., Ben92 CGA Far88 0.2 JM92 FSZ93 Ben9 GA 0.0 0 2 4 Q2 (GeV/C)2 6 8 Figure 514: The measured transparency for C, compared with several Glauber calculations. The references denoted by the legend are as follows: (i) KYS93 = Kohama et al. of ref. 66]; calculations using the reaction (o,,) and total (ut.t) free NN cross-sections are shown (ii) Nik9 = Nikolaev et al. of ref. 671, which gives the same result as KYS93 (iii) Ben9 = Benhar et al. of ref. 68]; GA = glauber approximation, CGA includes the correlation hole effect (iv) Far8 = Farrar et al. of ref. 28] (v) JM92 = Jennings and Miller of ref. 69] (vi) FSZ93 = Frankfurt, Strikman, and Zhalov of ref. 70] Ur 180 I I I 0. - --- ----- I I I KYS93a / Nk93 Ben92 CGA FSZ93 Far88 0.6 - - - - - - KYS93 att Ben92 GA I-' N '11 0 _ 04 - LL --- ---------- :Tt__ ___ -- ----- --- T------- - - - - -------- -------- ---------------- ---------------------- - - - - - - - - 0.2 0.0 I 0 I I 2 I I I 4 Q2 I (GeV/c )2 I I I 6 I I 8 Figure 15: The measured transparency for "Fe, compared with several Glauber calculations. Te notation in the legend is the same as that of Figure 514. 181 I .8 I JM92 KYS93 a, Ber92 CGA .6 I A- Nik93 FSZ93 Far88 Ben92 GA KYS93 atot I--, CN 4- I-- -- - - - - - - --- - - - - - - - - - - - --- - - - - - - - - - -- -- - - - -- -- -- -- -- - - -- -- - - - - - - - -- - - - - - - - ----- ------------------------------------------------- .2 .0 0 2 4 Q2 ( eV/c )2 6 8 Figure 516: The measured transparency for "'Au, compared with several Glauber calculations. The notation in the legend is the same as that of Figure 514. 182 70% that of the total cross-section in the high energy limit, and KYS93 and Nikolaev et al. of ref. 64] indicate that the resulting difference in transparency for carbon is about 0.1. This is a large effect, and should be evaluated carefully by computing the fraction of the elastically rescattered protons which ends up inside the experimental acceptance. Precise information about the acceptance is thus required. KYS93 points out that the strong forward peaking of the pN reaction will tend to make the reaction cross-section more appropriate. Ref. 641has used approximate values for the NE18 momentum cutoff to evaluate tis effect, and finds the resulting transparency to be midway between the two extreme cases. KYS93 further points out that the dependence of the cross-section on the relative momentum should be taken into account, and evaluate a Fermi smearing correction. his is found to be small (5%) in the case of 'He, but can be expected to increase for eavier nuclei and for the restrictive p. acceptance of NE18. The effect of short range correlations is also variously included in the theories. Benhar et al. (Ben92) 68] suggest that the short-range NN repulsion causes a "correlation hole" around the sruck nucleon. This is evaluated using an expression of the form T z Id 2bdz'p(r') e, f "' dz11aG(z11 -zl)p(rl) (5.8) Here the initial position of the struck proton is described by an impact parameter b and position z' aong its momentum vector; the integral over z" describes its path through the nucleus. The correlation function G(z - z) is negative and effectively clears a space around the travelling nucleon, reducing its interaction with the medium. The magnitude of this effect is found to be quite large: in Figure 517 the carbon transparency is shown to increase y 0.1. However, very different effects of correlations have been found by other workers. Nikolaev et al. (Nik93) 671point out that the so-called "spectator effect" largely cancels the "hole effect". Simply, the spectator effect accounts for the fact that 183 a "hole" in the nucleon density at one location must be balanced by a bunching of nucleons elsewhere. The effect is described by replacing the negative function G(z" - z) in Equation 5.8 with the expression (I + 1,,,,op(r")) in terms of the (positive) correlation length Figure A net increase of 2 18 shows the magnitude of the two effects found by this calculation. is determined, a result which is consistent with the independent calculation of KYS93. However, the magnitude of the individual effects is less than 5%, and so inconsistent with the - 20% hole effect of Ben92. The expressions found for the hole effect in Ben92 and Nik93 are identical, and so the discrepancy must come from the input to the calculation. Possible explanations are the nuclear density distributions employed, and the use of the total pN cross-section (Ben92) rather than the reaction cross-section (Nik93 and KYS93). The inclusion of Pauli blocking and effective mass corrections was shown (Figure 1-10) to contribute substantially to the transmissionat Q < 04 eV/c)'. These corrections are expected to be less important at NE18 energies 72]. One evaluation of Pauli blocking in the (GeV/c)' range is found in Ben92 for iron. As shown in Figure 517 the iron transparency is found to increase by 005 (- 12%) a non-negligible effect. 5.4.2 Comparisonwith ColourlYansparency Predictions The previous section demonstrated that there is no firm theoretical baseline for the nuclear transparency expected in the absence of PCD effects. A large range in magnitude is covered by the Glauber calculations, leaving us with only the Q-dependence of the data to provide a clear signature of new effects. However, the few Glauber estimates which incorporate the Q dependence of the free pN cross-section demonstrate that a substantial rise of T with momentum transfer can be expected in the absence of colour transparency as well (- 10% for 12C over the 184 Q2 = 3 to 7 range). One can thus expect F_ :1% Q C 0I.. aCL In C 0 .t: 'a F0 1% 2. *+ V o2((GeV/c)?] a IV Figure 517: Transparency calculations of Benhar et al. (Ben92) 681. GA and CA denote the Glauber and correlated Glauber approximations respectively, CT indicates the inclusion of colour transparency effects. For iron, two additional versions of the CGA + CT prediction are included: the dash-double-dotted line approximates the correlation function G(r') from Equation 5.8 with the pair correlation function; the dot-double-dashed line includes Pauli blocking. 185 I CZ _ I 10 &A %_--I c0 5- 0 0L_ 0- 0 U .......... -5- - 0- - 1.c - hole effect -------------------------- 0 net en ect ..................................................... ....................... ................. spectato'i " I 50 ____ 100 150 200 atomic number I I I 250 Figure 518: The correlation "hole" and "spectator" effects, as evaluated by Nikolaev et al. (Nik93) 671. that the interpretation of the data in terms of colour transparency will be ambiguous at best. The NE18 transparency measurements are compared in Figures 519 through 521 with a representative selection of calculations including colour transparency. The estimated size of these effects covers a very large range, and it is at once clear that only the most extreme prescriptions can be excluded by the data. Again we focus on the carbon data, for simplicity. The most drastic rise with Q' is found in the naive parton model of Farrar et al. ar88) 281 which was discussed in Section 1.5; the indicated > 40% rise over the NE18 kinematic range is clearly not observed. The quantum diffusion model from the same paper, however, predicts a more gradual rise of - 20%. The Benhar et al. (Ben92) 681CGA+CT calculation uses the same quantum diffusion prescription for the evolution of the struck hadron as it leaves the nucleus, but incorporates the correlation 186 hole effect described in the previous section. A 25% rise is indicated. Note that the Ben92 GA+CT calculation shown in Figure 517 indicates a larger rise of 30%. This can be understood by recalling the large size of the correlation hole effect found by these authors, and onsidering its effect on the colour transparency picture. The hadron of reduced size and interactions postulated by CT is now surrounded by a correlation hole, and so the early stages of its expansion go "unnoticed" by the nucleus. The difference between FSI with and without CT is thus reduced by the hole effect. Noteworthy also is the discrepancy between the 20% rise of the quantum diffusion model of Far88 and the 30% rise of the GA+CT curve, although these two calculations use the same CT formalism. The results are clearly dependent on the details of the calculation, particularly the input parameters such as the nuclear density distribution and in-medium pN cross-section. The other curves depicted in Figure 519 are those of Jennings and Miller (JM92) 69] and Frankfurt, Strikman, and Zhalov (FSZ93) 70]. Both are based on an expansion of the struck nucleon wavefunction in terms of excited hadronic basis states. JN192 is an extension of an earlier calculation 711,evaluated using only one such excited state. In JM92 an expansion over a spectrum of excited nucleons N* is considered; the density of these immediate states is parametrized by a function g(M2) where M is the mass of the N*. Various choices for the form of this function are considered, based on measurements of the N* spectrum from inelastic scattering and diffractive dissociation experiments. Two extreme CT predictions are provided by a power law form g(M2) (indicated by - M.,,O in Figure 519), and a distribution with a sharp mass cutoff. The former version determines a relatively weak CT effect 10% rise of NE18 Q2 range). The latter suggests a larger rise, and gives similar results to the calculation of FSZ93 - The FSZ93 calculation is distinguished by its use of an energy-dependent above Q2 of 3. free pN cross-section, and so is the only one which exhibits a rise at low 187 Q2. Deciding which prescriptions can be excluded by the data is an exercise of limited use, given the uncertainty in the "standard" FSI prescriptions on which the CT calculations are based. By shifting the calculations up or down by the 10% indicated by Figure 514, one can bring almost any CT curve into reasonable agreement with the data. It is clear that higher precision experiments in the Q > (GeV/c)' range are required to settle these questions and provide a sharper baseline for quasielastic scattering in this range. Only with more a exact understanding of "traditional" nuclear physics effects can one hope to look effectively for the signature of a new process. 5.4.3 A Dependence As pointed out in Section 1.5, one of the assumptions involved in the colour transparency effect is that the PLC produced at the scattering vertex expands to the size of a dressed proton over a distance at least as large as the nuclear radius. CT effects may therefore show up in the A dependence of the transparency as well as the Q dependence. Once again, the critical question is what to compare to: are CT effects required to explain the A dependence of the data? To answer this question, we invoke a simple model of the transmission probability. Classically, if a particle travels a distance x through a medium of containing p scatterers per volume and interacts with these scatterers via a cross-section O',the probability that the particle escapes without any interaction is P,.c., = "'. The combination Pa (5.9) is the mean free path, A. If one then considers the knockout of a proton from a spherical nucleus of radius R and uniform nucleon density po, assuming that the scattering probe samples the entire nuclear volume, one obtains the following 188 1.0 0.8 0.6 0 I--I 0.4 0.2 0.0 Q2 (GeV/ C)2 Figure 519: The measured transparency for 12C, compared with several calculations including colour transparency. The references denoted by the legend are as follows: (i) Far8 = Farrar et al. of ref. 28]; calculations in the naive parton and quantum diffusion models are sown; (ii) Ben9 = Benhar et al. of ref. 68]; CGA+CT indicates inclusion of the correlation hole effect (iii) JM92 = Jennings and Miller of ref. 69]; two forms of the density-of-states function gM,,,) (described in the text) are used (iv) FSZ93 = Frankfurt, Strikman, and Zhalov of ref. 70]. 189 I I I I I I I I I 0. - I I -------- I I I I I I I I T I I Ben92 CGA+CT FSZ 0.6 0 C-4 0 1-1 - 0.4 - L -- 0.2 0.0 I 0 IIIIII1 2 I I Q2 I 4 (GeV/c I I I 1 6 I 8 )2 Figure 520: The measured transparency for "Fe, compared with several calculations including colour transparency. The notation in the legend is the same as that of Figure 5-19. 190 I Q .U I I I I I t ---- . 1-1 I - - - - - JM92 -------- Ben92 CGA+CT JM92 g(M.2 _ M.-,$ - I FSZ93 .6 - CN g(M CUt I 4- I IIIIIIII .0 0 IIIIIIII 4 2 Q2 (GeV/c III 6 8 2 Figure 521: The measured transparency for 197 Au, compared with several calculations including colour transparency. The notation in the legend is the same as that of Figure 5-19. 191 expression for, the transmission probability: T= 3 8'3 with F R. A (2k - 1) + e-"4(2f + 1) Given the rms charge radius R,,,(A) (5-10) from previous data 73], the effective radius for a sphere of the same charge but uniform density is R number density of nucleons is simply po = 'A 4xR3 [5 5R,,,. 3 The uniform The values used for the nuclei studied by NE18 are given in table 5.5. The remaining unknown is the effective pN cross-section in the nucleus o,f f, and can be fit to the data at each Q2 using T(A). Only A > 2 are used in the fit as this model is clearly inaccurate for the single spectator nucleon of 2H. The fits to the data as a function of A are shown in Figure 522; the resulting values for ff are presented in table 56, along with the X2 per degree of freedom for the fit. The X values are all less than 1, indicating an excellent fit. One finds, then, that the classical attenuation model provides a reasonable parametrization of the data. Note that in the limit of complete colour transparency, o,ff --+ 0. Thus, the parameter aq f provides another measure of the nuclear transparency as a function of Q2, one that conveniently takes into account the data taken at all A. The fit values of 0,ff are also compared to the free pN cross-section in table 56. J obtained by averaging the free pp and pn cross-sections of ref. [5], weighted respectively with the number of protons (Z-1) and neutrons (N) in the spectator system. One sees that the 0',ff is about 60% lower thandfr,, at all Q2. Such a reduction could be expected from quantum effects not accounted for in the classical calculation, as well as nuclear effects such as Pauli blocking, short-range correlations, etc. We note that ref. 74] uses a similar classical model to parametrize nuclear transmission data at Q2 = 034 (GeV/c)', and also finds that the effective cross-section is substantially lower than 192 They parametrize o,,ff Table 5.5: Input values used in classical model of transmission. Target 12C Fe 197 Au 56 fm R (equiv. uniform radius) fm 2.50 3.76 5.33 3.23 4.85 6.88 Rrm's Table 56: Fit values of aq f using classical model of Equation 5.10. The isospin-average of the free pN cross-section is included for comparison. Oleff mb x 2 /deg. freedom 1 3 5 18 ± 2 27 ± 3 28 ± 3 0.7 0.4 0.6 34.9 43.0 41.8 0.56 ± 06 0.62 ± 07 0.67 ± 07 7 23 ± 3 0.1 41.4 0.5 Q2 (GeV/c)2 df ree a-ff dfee mb ± 10 in terms of an assumed density dependence: 17eff (p, r = I O'free _ Kp(r)' (5.11) Here, K rather than aq f itself) is the free parameter. This form is based on the work of ref. 75], which evaluates some of the nuclear effects mentioned above and arrives at a substantial dnsity dependent reduction of in-medium cross secion. 193 - L___] . I I I I I I I I I I'll I . . . I - I I id 02 = 1 (GeV/C)2 _ I I I 1 11.1 I I I . . .I. I = 5 (GeV/c)' Q 1 I I_'I 0.1 n . I I . I I I I I . . r-. 0. 1 10,- lo- loj I,, 10" , . I .- II I I I I 10j A . I . . I II . = 3 (2 I 104 A I . . . . I . . . . . GeV/C)2 1 , , - 02 - I . I . 1.111 11 , . = 6.8 (GeV/C)2 - 1-1 F__ nV. I 11 I II I . I . . 1.112 10 lou A I - - -1 0. lo" 1 I lo" I . . I .11- lo' . I II - lo" A Figure 522: NE18 measured transparencies for A> 2 targets. The solid lines indicate a one-parameter fit (at each Q2) to the classical transmission model described in the text. 194 Chapter 6 Discussion of Results Experiment NE18 has measured the coincidence quasielastic cross-section for (ee'p) scattering from te nuclei 11, '11, 12 c, 6 Fe, and 197 Au in a Q range of to 68 (GeV/C)2. This extends the previous momentum transfer range of such measurements by over an order of magnitude. The elastic hydrogen data were found to be well explained by the dipole form for the proton electric form factor, the parametrization of Gari and Kriimpelmann 41] for the proton magnetic form factor, and a radiative corrections prescription based on the work of Mo nd Tsai 58][591. This prescription included radiation by the relativistic knockout proton, and consisted of a reevaluation of the Mo and Tsai formulae in a coincidence framework. The measured hydrogen E distribution provided a precise test of the radiative cross section, and was found to be in excellent agreement with the calculation. Data analysis was performed entirely in a traditional nuclear physics framework, consisting of the Plane Wave Impulse Approximation (PWIA), the Independent Particle Shell Model (IPSM) of the nuclear structure, and the deForest off-shell ep cross-section [3]. These elements were programmed into a Monte Carlo simulation, which folded the 195 PWIA/IPSM cross-section with the experimental acceptances and the radiative crosssection; the data were then compared to this calculation. A momentum distribution determined using the Bonn potential was provided as input to the calculation for 2 H, and found to produce excellent agreement with the data. The IPSM spectral functions used for the heavier nuclei were based on fits to measurements made at lower momentum transfers. Certain parameters were also varied to provide a better fit to the NE18 Q2 = I data, and the effect of this variation on the results was included in the experimental systematics as a model dependent uncertainty. The data were found to be in good agreement with the calculation. The momentum distributions for the individual shells of the 12C Up to Q2 nucleus were separated, providing direct evidence of the nuclear shell model of 7 (GeV/C)2 . Furthemore, no evidence was found for additional or missing strength in the nuclear energy and momentum distributions. This is in contrast with measurements made at Q2 - 02 (GeV/c)': DWIA/CDWIA calculations of initial and final state distortions must be made to explain the data 6 and evidence of scattering from multinucleon currents has been observed at high E[10][12]. The primary quantitative result of the experiment was the measurement of the nuclear transparency T over a sizable range in Q2 and A. A classical model of the transmission probability was found to provide a good parametrization of the A-dependence of the results. This parametrization yielded effective values for the in-medium pN cross-section about 60% as large as the free cross-section. The Q2 dependence of T showed little evi- dence of colour transparency (CT) effects, expected to show up as a rise in transmission relative to the "standard" expectation. The data are consistent with a transparency above Q2 Q2 independent = 3 GeV/C)2, and extreme models of CT predicting rises of 30 - 40% over the NE18 kinematic range can be ruled out. However, the value of T expected in a conventional nuclear physics picture is not well established; values differing by - 196 20% have been suggested for the carbon nucleus, for example. Furthermore, the Q dependence of T in the absence of CT is not flat over the NE18 Q range due to the energy dependence of the pN cross-section. This Glauber calculations at Q > I Q2 dependence is not well known, for most have been carried out using a constant value for pN, equal to the asymptotic value reached in the high energy limit. These uncertainties cause the colour transparency signature to be obscured, and make further interpretation of the data in terms of CT effects ambiguous. The NE18 results have instead provided a baseline with which to constrain models of quasielastic scattering above Q = I GeV/C)2. Whether or not CT effects are involved, the NE18 results cleary demonstrate that PQCD effects are not, large in quasielastic scattering up to momentum transfers of 7 (GeV/ C)2. The surprise is then that the constituent counting rules of Section 14.2, also a result of PQCD, have proved so successful at describing form factor data and other exclusive measurements at similar kinematics. The arguments of ref. 241, demonstrating alternative non-perturbative means of accounting for the magnitude of the data, present a possible explanation and suggest that the momentum transfer threshold at which PQCD can validly b applied is much higher. The most striking result of this experiment has been the demonstration that our understanding of quasielastic scattering in a conventional nuclear physics picture extends over two orders of magnitude in momentum transfer. Further analysis of the data in this picture is possible. For example, the radiative Monte Carlo calculation provides us with the means to deradiate the extracted spectral functions in a largely model-independent way. The transparency for individual shells in 12C could then be determined, for ex- ample, providing a measure of the importance of density dependent effects and Fermi smearing in FSI calculations. Also, analysis of the asymmetry of the deuterium momentum distributions is underway 381. This asymmetry is sensitive to the structure 197 function WI of Equation 19. Several calculations of this structure function exist, taking various approaches to the difficult question of relativistic corrections. The future of these measurements clearly lies in higher precision and higher energy experiments. The quasielastic analysis performed for NE18 provides a good basis for the analysis of coincidence experiments at CEBAF, for example, where high intensity beams will be enable high statistics data taking in a similar Q range. Our results also demonstrate that to observe sizable colour transparency effects one must proceed to higher momentum transfers; for the foreseeable future, an investigation of this nature can be made only at SLAC. NE18 establishes the necessary baseline to which such higher energy measurements can be compared. 198 Bibliography [1] P. E. Bosted et al., Phys. Rev. Lett. 68, 3841 1992); A. Lung et al., Phys. Rev. Lett. 70, 718 1993). [21 R. G. Arnold et al., Phys. Rev. Lett., 57, 174 1986). [3] T. de Forest Jr., Nucl. Phys. A132, 305 1969). [4] S. Frullani and J. Mougey, Advances in Nucl. Phys. 14, Plenum Press, New York 1984. [5] "Particle Properties Data Booklet", ed. by M. Aguilar-Benitez et al., North-Holland (April 1992); Particle Data Group, Phys. Lett. [6] L. Lapiks, Nucl. Phys. A553, 297c 1993). [7] P. Barreau et al., Nucl. Phys 239 1990). A 402, 515 1983). [8] J. W. Van Orden and T. W. Donnelly, Ann. Phys. 131, 451 1980); T. W. Donnelly, J. W. Van Orden, T. deForest, Jr., and W. C. Herman, Phys. Lett. 76B, 393 1978). [91 J. M. Finn, R. W Lourie, and B. H. Cottman, Phys Rev. C 29, 2230 (1984). [10] R. W. Lourie et al., Phys. Rev. Lett. 56, 2364 1986). [11] J. Mougey et al., Nucl. Phys. A262, 461 1976). 199 [12] P. E. Ulmer et al., Phys. Rev. Lett. 59, 2259 1987). [13] R. J Glauber, "Lectures in theoretical physics", vol. 1, ed. W. E. Brittin and D. G. Dunhan, Interscience 1959). [14] G. Garino et al., Phys. Rev. C 45, 780 1992). [15] G. R Farrar, Phys. Rev. Lett, 53, 28 1984). [16] G. P Lepage and S. J. Brodsky, Phys. Rev. D 22, 2157 1980). [17] S. J. Brodsky and G. R. Farrar, Phys. Rev. Lett. 31, 1153 1973); S. J. Brodsky and G. R. Farrar, Phys. Rev. D 11, 1309 1975). [18] P. V. Landshoff, Phys. Rev. D 10, 1024 1974). [191 S. J. Brodsky and G. P. Lepage, Phys. Rev. D 24 2848 1981). [20] S. J. Brodsky and G. F. de Teramond, Phys. Rev. Lett. 60, 1924 1988). [21] C. W. Akerlof et al., Phys. Rev. 159, 1138 1967). [22] S. J Brodsky and G. F. de Teramond, Phys. Rev. Lett. 60, 1924 (1988). [23] V. L. Chernyak, A. A. Ogloblin, and 1. R. Zhitnitsky, Novosibirsk preprints INP 87-135,136; 1. D. King and C. T. Sachrajda, Nucl. Phys. B 279, 785 and N. Stefanis, Phys. Lett. 987) M Gari 187, 401 (1987). [24] N. Isgur and C. H. Llewellyn Smith, Phys. Rev. Lett. 52, 1080 1984); N. Isgur and C. H. Llewellyn Smith, Nuc. Phys. B317, 526 (1989). [25] L. Frankfurt, G. A. Miller, M. Strikman, Comments Nucl. Part. Phys., 21, 1 (1992). [26] D. Perkins, Phil. Mag. 46, 1146 1955). 200 [27] B. Povh, J. Hfifner, Phys. Rev. Lett. 58, 1612 1987). [28] G. Farrar, H. Liu, L. L. Frankfurt, and M. 1. Strikman, Phys. Rev. Lett. 61, 686 (1988). [29] L. L. Frankfurt and M. I. Strikman, Phys. Rep. 160, 235 1988). [301 A. S. Carroll et al., Phys. Rev. Lett. 61, 1698 1988). [31] J. P. Ralston and B. Pire, Phys. Rev. Lett. 61, 1823 1988). [32] D. H. Potterveld, Ph. D. Thesis, Caltech. [33] Y. S. Tsai, Rev. Mod. Phys. 46, 815 1974). [34] R. Anderson et al., Nucl. Instr. Meth. 66, 328 1968). [351 P. N. Kirk et al., Phys. Rev. D8, 63 1973). [361 W. R. Leo, "Techniques for Nuclear and Particle Physics Experiments", SpringerVerlag 1992). [37] K. L. Brown et al., SLAC Report No. 91, Rev. 2 1977. [38] H. J. Bulten, private communication. [39] J. T. O'Brien, H. Crannel, et al., Phys. Rev. C 9 1418 1974). [40] T. C. O'Neill, Ph. D. thesis, Caltech 1994). [41] M. Gari and W. Kriimpelmann, Z. Phys. A322, 689 1985). [42] C. Giusti and F. D. Pacati, Nucl. Phys Pacati, Nucl. Phys A 485, 461 1988). 201 A 473, 717 1987); C. Giusti and F D. [43] K. Holinde and R. Machleidt, Nucl. Phys A 256, 479 1976). [44] F. C. Perey and B. Buck, Nucl. Phys. 32, 353 1962); H. Fiedelday, Nucl. Phys. 77, 149 1966); H. P. Blok and J. H. Heisenberg, from "Computational Nuclear Physics, vol. 1", ed. by W. E. Brittin et al., Interscience 1959). [45] C. E. Brown and M. Rho, Nucl. Phys A 372, 397 1981). [46] E. N. M. Quint, Ph. D. Thesis, U. Amsterdam 1988). [47] J. Mougey et al.. Nucl. Phys A 262, 461 1976). [48] J. W. Negele, Phys. Rev. C1, 1260 1970); J. W. Negele and D. Vautherin, Phys. Rev. C5, 1472 1972). [49] L. Lapikas, private communication. [50] G. Jacob and Th. A. J. Maris, Rev. Mod. Phys. 45 6 1973). [51] J. W. Van Orden, W. Truex, and M. K. Banerjee, Phys. Rev. C21, 2628 (1980). [52] X. Ji, private communication. [53] 1. Sick, private communication. [54] S. Liuti, private communication. [55] W. Atwood, Ph. D. thesis, SLAC-Report [56] G. G. Petratos, no. 185 1975). private communication. [571 D. Wasson et al., to be published. [58] L. M. Mo and Y. S. Tsai, Rev. Mod. Phys. 41, 205 1969). 202 [59] Y. S. Tsai, Phys. Rev. 122, 1898 1961). [601 C. de Calan, H. Navelet, and J. Picard, Nucl. Phys. B348, 47 1991). [61] D. R Yennie, S. Frautschi, and H. Suura, Ann. Phys. (N.Y.) 13, 379 (1961). [62] N. C. R. Makins et al., Phys. Rev. Lett. 72, 1986 1994). [63] T. C. O'Neill et al., submitted to Phys. Rev. Lett. [64] N. N. Nikolaev, private communication. [65] D. Wasson, private communication. [66] A. Kohama, K. Yazaki, and R. Seki, Nucl. Phys A 551, 687 1993). [67] N. N. Nikolaev et al., Phys. Lett. 317, 281 1993). [68] 0. Benhar et al., Phys. Rev. Lett. 69, 881 1992). [69] B. K. Jennings and G. A. Miller, Phys. Rev. Lett. 69, 3619 1992). [70] L. Frankfurt, M. Strikman, and M. Zhalov, to be published. [71] B. K. Jennings and G. A. Miller, Phys. Rev. D 44, 692 1991). [72] S. Pieper, private communication. [73] "Structure of the Nucleus", M A Preston and R. K. Bhaduri, Addison-Wesley (1975). [74] D. F. Geesaman et al., Phys. Rev. Lett. 63, 734 1989). [751 V. R. Pandharipande and Steven C. Pieper, Phys. Rev. C 45, 791 1992). 203