University of California

Los Angeles

Creation, Manipulation, and Diagnosis of Intense,

Relativistic Picosecond Photo-electron Beams

A dissertation submitted in partial satisfaction

of the requirements for the degree

Doctor of Philosophy in Physics

by

Scott Geoffrey Anderson

2002
 © Copyright by

Scott Geoffrey Anderson

2002
The dissertation of Scott Geoffrey Anderson is approved.

Alfred Wong

Chandrashekhar Joshi

Claudio Pellegrini

James B. Rosenzweig, Committee Chair

University of California, Los Angeles

2002

ii
To my parents…

iii
 Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . 1
1.1 Emittance and Brightness . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.1 Phase Space Distributions . . . . . . . . . . . . . . . . . . . . 2

1.1.2 Moments of the Distribution and Emittance . . . . . . . . . 7

1.1.3 Emittance and Brightness as Figures of Merit . . . . . . . . 12

1.2 Photoinjector Development . . . . . . . . . . . . . . . . . . . . . . . 16

1.2.1 DC Guns and Thermionic Cathodes . . . . . . . . . . . . . . 17

1.2.2 RF Cavities and Photo-cathodes . . . . . . . . . . . . . . . . 21

1.2.3 Photoinjectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.3 The Neptune Photoinjector . . . . . . . . . . . . . . . . . . . . . . . 25

1.3.1 The Neptune Advanced Accelerator Laboratory . . . . . . . 26

1.3.2 Accelerating Sections . . . . . . . . . . . . . . . . . . . . . . . 28

1.3.3 Photo-cathode Drive Laser . . . . . . . . . . . . . . . . . . . . 34

1.3.4 RF System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

1.3.5 Magnetic Compressor . . . . . . . . . . . . . . . . . . . . . . . 41

1.3.6 Beam Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . 47

1.3.7 Control System . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

1.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

iv
2 Rectilinear Space-charge Dynamics . . . . . . . . 54
2.1 Linear Space-charge Emittance Compensation . . . . . . . . . . . . 57

2.1.1 Space-Charge Dominated Behavior of Beam Slices . . . . . 57

2.1.2 Envelope Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 70

2.1.3 Brillouin Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

2.1.4 The Invariant Envelope . . . . . . . . . . . . . . . . . . . . . . 80

2.2 Space-Charge Dynamics within a Slice . . . . . . . . . . . . . . . . . 84

2.2.1 Slab Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

2.2.2 Cylindrically Symmetric Beams . . . . . . . . . . . . . . . . . 97

2.2.3 Simulation of Coasting Cylindrical Beams . . . . . . . . . . 101

2.2.4 Accelerating Cylindrical Beams . . . . . . . . . . . . . . . . . 109

2.3 Application to a Real System . . . . . . . . . . . . . . . . . . . . . . 117

3 Emittance Measurement of Photoinjector Beams 130

3.1 Emittance Measurement Techniques . . . . . . . . . . . . . . . . . . 132

3.1.1 Multislit-based Measurement . . . . . . . . . . . . . . . . . . 132

3.1.2 Quadrupole Scanning Emittance Measurements . . . . . . 138

3.2 Experimental Setup and Procedure . . . . . . . . . . . . . . . . . . . 142

3.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

3.4 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

v
4 Magnetic Compression

and Emittance Growth . . . . . . . . . . . . . . . . . 162

4.1 Experimental Measurements at Neptune . . . . . . . . . . . . . . . 166

4.1.1 Pulse Length Measurements . . . . . . . . . . . . . . . . . . . 171

4.1.2 Emittance Measurements . . . . . . . . . . . . . . . . . . . . 180

4.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

4.2.1 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

4.2.2 A Heuristic Model . . . . . . . . . . . . . . . . . . . . . . . . . 194

4.2.3 Bifurcations and Size Effects in Simulation . . . . . . . . . 205

4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

A Photoinjector Simulations . . . . . . . . . . . . . 216

A.1 parmela . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

A.2 norse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

A.3 bender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

References . . . . . . . . . . . . . . . . . . . . . . . . 225

vi
 Figures

1.1 The bi-Gaussian phase space distribution function . . . . . . . . . 6

1.2 The effect of nonlinear forces on the distribution function. . . . 8

1.3 The rms ellipse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.4 A Pierce type DC gun . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

1.5 A single cell rf gun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.6 Example Photoinjector Gun . . . . . . . . . . . . . . . . . . . . . . . 24

1.7 The Neptune Advanced Accelerator Laboratory . . . . . . . . . . . 27

1.8 The Neptune Photoinjector . . . . . . . . . . . . . . . . . . . . . . . 29

1.9 1.6 cell gun mode separation measurement . . . . . . . . . . . . . 32

1.10 Photoinjector drive laser . . . . . . . . . . . . . . . . . . . . . . . . . 35

1.11 The Neptune RF system . . . . . . . . . . . . . . . . . . . . . . . . . 38

1.12 Single-shot PWT phase measurement . . . . . . . . . . . . . . . . . 41

1.13 The Neptune chicane compressor . . . . . . . . . . . . . . . . . . . 42

2.1 Space-charge expansion of beam slices . . . . . . . . . . . . . . . . 62

2.2 Beam Slices after Thin Lens . . . . . . . . . . . . . . . . . . . . . . . 64

2.3 Waist and cross-over slice trajectories . . . . . . . . . . . . . . . . 65

2.4 Plasma wavenumber versus cross-over focal strength . . . . . . . 66

2.5 The emittance compensation model . . . . . . . . . . . . . . . . . . 69


2.6 Trace space trajectories for σr , σr
in a uniform focusing channel 77

vii


2.7 Projected trace space areas for σr , σr
in a uniform focusing

channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

2.8 Projected trace space area in a uniform focusing channel with

σr 0 < σeq . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

2.9 Envelope oscillations about their invariant envelopes . . . . . . . 83

2.10 Evolution of a freely expanding slab beam distribution . . . . . . 89

2.11 Trace space of a slab symmetric beam at wave-breaking . . . . . 94

2.12 Matched versus mismatched slab symmetric beam density pro-

files at wave-breaking . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

2.13 A comparison of the function g (r0 ) with Gaussian f (r0 ) . . . . 100

2.14 Simulation of a freely expanding, cylindrically symmetric, ini-

tially Gaussian beam . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

2.15 Phase space of a freely expanding, cylindrically symmetric, ini-

tially Gaussian beam . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

2.16 Simulation of a cylindrically symmetric, initially parabolic beam

through one thin lens . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

2.17 Simulation of a cylindrically symmetric, initially parabolic beam

through many thin lenses . . . . . . . . . . . . . . . . . . . . . . . . 105

2.18 Evolution of the beam distribution transported through a thin

lens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

2.19 Evolution of the emittance for a beam rms matched to a uniform

focusing channel from simulation and analytical prediction . . . 109

viii
2.20 Evolution of the emittance for a beam rms matched to the in-

variant envelope from simulation and analytical prediction . . . 113

2.21 Trace space of an initially parabolic beam slice at the maximum

emittance point in the accelerating beam simulation. . . . . . . . 114

2.22 The beam envelope evolution for the same simulation as Fig-

ure 2.20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

2.23 The beam envelope and emittance evolution for the LCLS work-

ing point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

2.24 Phase and configuration space simulations at the first emittance

compensation point, uniform case . . . . . . . . . . . . . . . . . . . 119

2.25 Phase and configuration space simulations at the end of the

simulation, uniform case . . . . . . . . . . . . . . . . . . . . . . . . . 120

2.26 The beam envelope and emittance evolution for the LCLS work-

ing point, longitudinal nonuniform case. . . . . . . . . . . . . . . . 123

2.27 Phase and configuration space simulations at the first emittance

compensation point, longitudinal nonuniform case . . . . . . . . 124

2.28 Phase and configuration space simulations at the end of the

simulation, longitudinal nonuniform case . . . . . . . . . . . . . . 124

2.29 The beam envelope and emittance evolution for the LCLS work-

ing point, radially nonuniform case. . . . . . . . . . . . . . . . . . . 125

2.30 Phase and configuration space simulations at the first emittance

compensation point, radially nonuniform case . . . . . . . . . . . 126

ix
2.31 Phase and configuration space simulations at the end of the

simulation, radially nonuniform case . . . . . . . . . . . . . . . . . 126

2.32 Phase space of the central beam slice for the uniform and radi-

ally nonuniform distributions . . . . . . . . . . . . . . . . . . . . . . 127

3.1 Illustration of the multislit-based emittance measurement scheme.132

3.2 Trace space reconstruction from emittance slits . . . . . . . . . . 135

3.3 The LLNL Thomson source photoinjector beamline. . . . . . . . . 143

3.4 Data from one quad scan emittance measurement . . . . . . . . . 148

3.5 Data from one slit-based emittance measurement . . . . . . . . . 148

3.6 Comparison of emittance measurements from the slit-based and

quad scan systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

3.7 parmela simulations of the emittance at the quadrupole and

the measured emittance using both measurement methods. . . . 150

3.8 homdyn simulation of the quadrupole scan plotted with the

parmela simulations and the measured emittance data. . . . . . 151

3.9 Path of the quad scan data in the (kp Ld , kp βx ) plane. . . . . . . . 154

3.10 Asymmetry in the quad scan data and simulations. . . . . . . . . 155

3.11 Simulations of a quad scan with and without emittance. . . . . . 156

3.12 Systematic errors in a quad scan of a space-charge dominated

beam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

4.1 Magnetic field in the chicane computed by amperes. . . . . . . . 168

4.2 LabVIEW program for chicane control. . . . . . . . . . . . . . . . . 170

x
4.3 Schematic of the CTR-based bunch length diagnostic. . . . . . . . 172

4.4 CTR-based pulse length measurements. . . . . . . . . . . . . . . . 178

4.5 Pulse length versus chicane current measurements. . . . . . . . . 179

4.6 Slit Image Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

4.7 Compressed beam emittance versus PWT Phase. . . . . . . . . . . 184

4.8 Trace Space Bifurcation. . . . . . . . . . . . . . . . . . . . . . . . . . 186

4.9 Compressed beam emittance versus PWT Phase and Chicane

Current. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

4.10 Compression-induced emittance growth as a function of beam

size in the chicane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

4.11 Slit collimated beam images for different beam sizes in the chi-

cane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

4.12 Simulations of compressed beam emittance versus PWT Phase. . 193

4.13 Simulated longitudinal phase space at the chicane entrance. . . . 195

4.14 Heuristic model motion in the chicane. . . . . . . . . . . . . . . . . 196

4.15 Horizontal dispersion in the chicane. . . . . . . . . . . . . . . . . . 197

4.16 The force between ellipsoidal slices. . . . . . . . . . . . . . . . . . . 201

4.17 Model calculation of the space-charge kick. . . . . . . . . . . . . . 201

4.18 Model calculation of the space-charge kick versus beam size. . . 203

4.19 bender simulation of the beam’s configuration space after com-

pression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

xi
4.20 bender simulation of the beam’s horizontal phase space after

compression. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

4.21 parmela simulation of the compressed beam’s longitudinal phase

space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

4.22 parmela simulation of the compressed beam’s configuration

space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

4.23 parmela simulation of the compressed beam’s energy distribu-

tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

4.24 tredi simulation of the compressed beam’s longitudinal phase

space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

4.25 tredi simulation of the compressed beam’s energy distribution. 210

4.26 bender simulation of emittance versus initial beam size. . . . . 211

A.1 The electric field of a uniformly charged ellipsoid. . . . . . . . . . 224

xii
 Tables

1.1 X-ray FEL electron beam requirements . . . . . . . . . . . . . . . . 15

2.1 Values of the form factor α for various slab-symmetric initial

beam distribution types . . . . . . . . . . . . . . . . . . . . . . . . . 91

2.2 Values of the form factor α for various cylindrically symmetric

initial beam distribution types . . . . . . . . . . . . . . . . . . . . . 101

2.3 Values of the form factor α for various cylindrically symmetric

initial beam distribution types, accelerating case. . . . . . . . . . 112

3.1 Electron beam parameter range used for emittance measure-

ments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

xiii
 Acknowledgments

There are many people I need to acknowledge for their support and con-

tribution to my graduate career. As it is impossible to list them all here, I

limit these brief words to those most directly responsible for the production

of this thesis.

First I thank my advisor, Jamie Rosenzweig, who leads by example, and

without whom this dissertation would not exist. Jamie is to thank for my

involvement in each of the experimental and theoretical topics in this thesis,

and his guidance and “calibrated provocations” always kept me moving on

the path I needed to follow, even if I didn’t always want to be. On the topic of

intellectual guidance, I also need to thank Claudio Pellegrini and Luca Serafini

for their help, support, and interest in my work in graduate school.

The experiments performed at Livermore were made possible by Greg

LeSage, and important work on the experiments was done by Greg, John

Crane, Gerry Anderson, and Bill Patterson. Thanks guys, and see you soon.

For the experiments at Neptune, a lot of people worked very hard to get

that photoinjector up and running. The list of people who got and kept the

machine working includes Pietro Musumeci, Matt Thompson, Hyyong Suk,

Travis Holden, Aaron Tremaine, Kip Bishofberger, Mike Loh, Soren Telfer,

Salime Boucher, Joel England, and Ron Augustsson. The work done by these

people was always important, but seldom recognized properly. I thank you

all. Of course, the photoinjector at Neptune exists within the larger context

of the laboratory, and I thank Chris Clayton and Chan Joshi for making that

xiv
possible.

Special thanks to Pietro Musumeci and Matt Thompson for their work

on the bunch compression experiments. In addition to helping perform the

experiments, Matt and Pietro each made critical components of the system

work. These experiments worked thanks to them. I should note that Pietro

was also very generous, always willing to offer me his antagonism.

I thank the Particle Beam Physics Lab, an entity that seems to have a life

of its own, and the lab members who came before me, Gil Travish, Mark

Hogan, Aaron Tremaine, Pedro Frigola, Sven Reiche, Ding Xiaodong, and Alex

Murokh, who taught me a lot about the nature of experimental physics that

isn’t written in books. The friendships and camaraderie of PBPL made the

long process of grad school bearable and even fun.

Finally, I thank my parents, who’s continuous support, and encourage-

ment has allowed me to succeed.

xv
 Vita

1973 Born, Edina, Minnesota.

1994 Undergraduate Internship, University of Minnesota Super-

computing Institute

1995 B.S., Physics, University of Minnesota

1995-1996 Teaching Assistant, Department of Physics, University of

California, Los Angeles

1996-2002 Graduate Research Assistant, UCLA Particle Beam Physics

Laboratory

1996 M.S., Physics, University of California, Los Angeles

1997 US Particle Accelerator School, Massachusetts Institute of

Technology

1998 US Particle Accelerator School, University of Texas, Austin

Presently Graduate Research Assistant, UCLA Particle Beam Physics

Laboratory

xvi
 Publications

S. Reiche, et. al., “Experimental Confirmation of Transverse Focusing and

Adiabatic Damping in a Standing Wave Linear Accelerator,” Phys. Rev. E, 56,

3572, (1997).

J. B. Rosenzweig, et. al., “Space Charge Dominated Envelope Dynamics of

Asymmetric Beams in RF Photoinjectors,” In Proc. 1997 Particle Accelerator

Conference, Vancouver, 1997, IEEE, 1965, (1997).

M. Hogan, et. al., “Measurements of High Gain and Noise Fluctuations in a

SASE Free Electron Laser,” Phys. Rev. Lett., 80, 289, (1998).

M. Hogan, et. al., “Measurements of Gain Larger than 105 at 12 µm in a Self-

Amplified Spontaneous-Emission Free-Electron Laser,” Phys. Rev. Lett.,81,

4867, (1998).

J. B. Rosenzweig, et. al., “The Neptune Photoinjector,” Nucl. Instrum. Methods

Phys. Res., Sect. A, 410, 437, (1998).

A. Tremaine, et. al., “Observation of Self-Amplified Spontaneous-Emission-

Induced Electron-Beam Microbunching Using Coherent Transition Radiation,”

Phys. Rev. Lett., 81, 5816, (1998).

xvii
S. Anderson, et. al., “Commissioning of the Neptune Photoinjector,” In Proc.

1999 Particle Accelerator Conference, New York, 1999, IEEE, 2006, (1999).

J. B. Rosenzweig, et. al., “A Comparison Between The Performance of Split

And Integrated RF Photoinjectors,” In Proc. 1999 Particle Accelerator Confer-

ence, New York, 1999, IEEE, 2039, (1999).

J. B. Rosenzweig, et. al., “The Effects of RF Asymmetries on Photoinjector

Beam Quality,” In Proc. 1999 Particle Accelerator Conference, New York,

1999, IEEE, 2042, (1999).

J. B. Rosenzweig, et. al., “Optimal Scaled Photoinjector Designs for FEL Ap-

plications,” In Proc. 1999 Particle Accelerator Conference, New York, 1999,

IEEE, 2045, (1999).

S. G. Anderson and J. B. Rosenzweig, “Non-equilibrium Transverse Motion

and Emittance Growth in Space-charge Dominated Beams,” Phys. Rev. ST

Accel. Beams, 3, 094201, (2000).

S. G. Anderson, et. al., “Commissioning of the Neptune Photoinjector,” In

Proc. 2001 Particle Accelerator Conference, Chicago, 2001, IEEE, 89, (2001).

S. G. Anderson, et. al., “Emittance Measurements of the Space Charge Dom-

inated Thomson Source Photoinjector,” In Proc. 2001 Particle Accelerator

Conference, Chicago, 2001, IEEE, 2260, (2001).

xviii
S. G. Anderson et. al., “Space-charge Effects in High Brightness Electron Beam

Emittance Measurements,” Phys. Rev. ST Accel. Beams, 5, 014201, (2002).

xix
 Abstract of the Dissertation

Creation, Manipulation, and Diagnosis of Intense,

Relativistic Picosecond Photo-electron Beams

by

Scott Geoffrey Anderson

Doctor of Philosophy in Physics

University of California, Los Angeles, 2002

Professor James B. Rosenzweig, Chair

The radio frequency photoinjector is the pre-eminent source for advanced

electron beam applications that require extremely high phase space den-

sity (high brightness) beams. Because of their high phase space density, the

collective fields generated by photoinjector beams dominate their behavior.

These space-charge fields influence every aspect of the beam’s handling, in-

cluding its acceleration, measurement, and transport. The effects of space-

charge must be carefully considered in all of these beam handling procedures

in order to deliver the highest brightness beams possible.

This dissertation investigates the space-charge dominated physical pro-

cesses involved in the acceleration and propagation, emittance measurement,

and magnetic compression of photoinjector beams. In the analysis of the

behavior of these beams, emphasis is placed on the techniques used to com-

pensate for space-charge forces, and maximize beam brightness.

The rectilinear motion of a space-charge dominated beam is analyzed,

xx
including both linear and nonlinear self forces, in order to determine the

evolution of the beam’s transverse emittance as it is accelerated and trans-

ported through the photoinjector. It is found that the emittance can be made

to oscillate by judicious use of external forces, and that this oscillation can

be manipulated to minimize the beam’s emittance, compensating for the ef-

fects of both linear and nonlinear space-charge forces, at a given location of

interest.

The creation of a high brightness beam in the presence of emittance os-

cillations is critically dependent on phase space diagnosis. Thus the mea-

surement of emittance of intense beams is investigated experimentally, the-

oretically, and in simulation, for quadrupole scanning and multi-slit based

measurement techniques. The quadrupole scanning method is found to have

systematic errors for space-charge dominated beams, and experimental mea-

surements using this technique give consistently higher emittance values

than both the slit-based measurements and simulations.

Finally, the measurement of emittance growth and transverse phase space

distortions induced by magnetic compression of the beam to sub-picosecond

lengths is described. A clear bifurcation of the phase space is observed when

the beam is strongly compressed. This effect is found to be correlated to the

folding of the beam distribution in configuration space.

xxi
 Chapter 1

Introduction

The radio frequency (rf) photoinjector embodies the current state of the art

in high quality electron beam generation. Its development was driven by

the needs of the accelerator user community, and these needs continue to

be a motivating factor in the refinement of these sources. As the quality

of beams produced with photoinjectors has improved and continues to im-

prove, the way we model, measure, and manipulate them has to change as

well. The main reason for this is that the nature of the forces that dominate

the beam’s evolution changes as its quality increases. Specifically, as the qual-

ity (an idea that will be quantified in the next section) of beams increases,

they become controlled by the self-fields, termed space-charge fields, that

they produce. This is in contrast to the classical model of an electron beam

as a set of non-interacting particles with random velocities determined by

a thermalized statistical distribution. This dissertation examines the ways

in which photoinjector beams are modeled theoretically, measured experi-

mentally, and manipulated to increase quality, in view of their space-charge

dominated behavior.

This chapter begins with a discussion of the concepts of emittance and

beam brightness and the importance of these ideas to the remainder of the

1
dissertation. The photoinjector is then introduced by reviewing the key tech-

nological advances that led to its invention and by showing how these ad-

vances increase the quality of the beam being produced. Finally, the pho-

toinjector in UCLA’s Neptune Laboratory, where much of the experimental

work relevant to this thesis was performed, is presented, and its component

parts are described.

1.1 Emittance and Brightness

The purpose of the photoinjector, and the goal in operating one is to pro-

duce the highest possible quality electron beam. It is therefore necessary to

specify exactly what is meant by the term “quality.” This is done by intro-

ducing the concepts of emittance and brightness. These terms arise through

a mathematical description of the beam. This description is summarized

below.

1.1.1 Phase Space Distributions

i )
An ensemble of particles is described by the three spatial coordinates (x

i ) of each of its N particles. We can

and three momentum components (p

 p)
represent the ensemble in a six-dimensional phase space (x,  as a collection

of points. This distribution of particles in phase space can be written as

 
N



 =
 p
f x, δ3 x−x −p
i δ3 p i , (1.1)
i=1



where δ3 x−x −p
i δ3 p i is a product of Dirac delta functions that local-

i , p
ize particle i in phase space as (x i ). This formulation of the distribution

2
is useful in simulations of beams; however, in order to analyze the ensemble

of particles theoretically, a smoothed distribution function is used.


 p
If we write the particle distribution function f x,  in phase space, then

the number of particles within a small volume dV is given by

dN = f dV = f d3 x d3 p. (1.2)

The use of a continuous function to describe the distribution is justified for

photoinjector beams as the total number of particles is typically on the order

of 109 or higher, and the number dN in any measurable volume dV is still

quite large. Similarly, the volume of phase space occupied by the beam is

V = d3 xi d3 pi . (1.3)

For real beams this volume will always be non-zero, because of thermal veloc-

ities at the source, nonlinear forces (both external and due to space-charge)

and other imperfections.


 p,
The goal of this analysis then, is to find the distribution function f x,  t

at a later time t, given an initial distribution and the forces involved in the

problem. If particles are not created or destroyed in the system, we can write

a continuity equation for phase space density:

∂f    
˙
 x · xf
+∇  +∇ ˙
 p · pf
 = 0, (1.4)
∂t
where the subscript of the divergence operators indicate differentiation with

 and p.
respect to x  If a Hamiltonian can be written for the particle motion,

then we may rewrite this equation using ∂H/∂p = dx/dt and ∂H/∂x =

−dp/dt as
 
∂f 3
∂ 2H ∂ 2H
+f − ˙
+x ˙
 x f + p
·∇ ·∇
 p f = 0. (1.5)
∂t j=1
∂x j ∂pj ∂pj ∂xj

3
The terms in the sum cancel, and what is left is the total time derivative of

the distribution function.

df ∂f ˙ ˙
 x f + p  p f = 0.
= +x
·∇ ·∇ (1.6)
dt ∂t

More compactly written, this is a statement of Liouville’s theorem:

df
= 0. (1.7)
dt

In words, this says that the density of particles in phase space does not

change, as one might expect, given that we started with an equation of con-

tinuity in this space. Liouville’s theorem may also be given in terms of the

volume of phase space occupied by the beam,


d
d3 xi d3 pi = 0. (1.8)
dt

Liouville’s theorem is an important result for the description of beams

and can be found in texts on classical mechanics (Symon [1] for example,

also derivations of Liouville’s theorem as it applies to beams can be found

in many accelerator physics texts such as Lawson [2] or Reiser [3].) This

result applies to distributions in six-dimensional phase space; however, if

the particle motion in one dimension (x) is independent of the other two

(y, z), then it also applies to the phase plane, defined by (x, px ),

Ax = dx dpx = constant. (1.9)

This is usually not the case with photoinjector beams, as the transverse

space-charge forces are strongly coupled to longitudinal position, as dis-

cussed in Chapter 2.

4
 Liouville’s theorem does not imply that the distribution function doesn’t

change its shape in time (or, of course, that individual particle’s position

and momentum do not change — they change according to the Hamiltonian.)

The evolution of the distribution function is given by Equation (1.6). An

example solution to this equation serves to illustrate the properties of the

distribution function. If we search for an equilibrium solution in the x phase

plane (assuming Equation (1.9) is valid) then Equation (1.6) reduces to

dx dfx dfx
+ Fx = 0. (1.10)
dt dx dpx

We may further simplify the problem for the purposes of this example by

assuming that the forces involved are linear in x and independent of px ,


and that the distribution function is separable (so we can write fx x, px =


X (x) P px [4]). In this case, Equation (1.10) has a seperable solution of the

form

 x2 p2
fx x, px = f0 exp − exp − x2 , (1.11)
2σx2 2σpx

where f0 is a normalization constant chosen so that the integral of fx over

the entire phase plane yields the total number of beam particles, N. The

widths (σx , σpx ) of the distribution are determined by the strength of the

linear restoring force, and the average kinetic energy (or temperature) of the

ensemble.

This distribution, termed bi-Gaussian, is shown in Figure 1.1a. The white

contour in the figure is a curve of constant density, and we can see from

Equation (1.11) that these curves are ellipses given by

x2 px2
+ = constant. (1.12)
σx2 σp2x

5
 3 3

2 2

1 1

px
px

p x/σ
0
p x/σ

-1 -1

-2 (a) -2 (b)
-3 -3
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
x/σ x/σ
x x

Figure 1.1: (a) Intensity map of the bi-Gaussian phase space distribution
function. (b) The distribution function after the restoring force is removed,
and the particles are allowed to drift.

The single particle dynamics of this system are revealed if we make the as-

sociations x 2 ∝ V and px2 ∝ T , so that Equation (1.12) states that the sin-

gle particle Hamiltonian (H = T + V ) is a constant of the motion — this is

expected since we have assumed a time-independent Hamiltonian. The con-

stant density ellipses correspond, therefore, to constant total energy of single

particles, and as such they trace the path in phase space that a particle moves

along in time. Note that this also implies that no particle in the distribution

may cross one of these curves.

In this example the distribution function was found by assuming an equi-

librium solution (∂f /∂t = 0.) If the single particle forces are latter changed

or removed, the distribution will evolve. In the simple case were the forces


are removed, the distribution at a later time, t, becomes fx x (t) , px (t) ,

6
where


px 0
x (t) = x0 + t, and
m


px (t) = px 0 . (1.13)

This case is illustrated in Figure 1.1b. We see from the figure that the curves

of constant density are again ellipses, although now there is a correlation

between x and px . The ellipses now change in time, but the property that

particles do not cross them still applies. The number of particles inside the

white curve in part b of the figure is the same as that of part a. Given this,

Liouville’s theorem says that the areas of the two ellipses are the same.

The fact that the area of an ellipse in phase space that contains a given

portion of the distribution remains constant if the forces involved are linear,

is very useful in beam physics, and is closely related to the emittance of a

beam, as defined in the next section. If the forces in the problem are not

linear, the initial ellipse of constant density will deform. This is illustrated in

Figure 1.2, where the bi-Gaussian distribution is exposed to nonlinear forces

(Fx ∝ sin (x).) In this case we see that the ellipse becomes filamented as

the distribution evolves. Notice here that Liouville’s theorem still applies

(the original ellipse and the final curve enclose the same area) but the quality

of the distribution is diminished; the smallest ellipse that encloses the final

distribution is much larger than the original.

1.1.2 Moments of the Distribution and Emittance

While the distribution function is a complete description of the ensemble

of particles, it is often not necessary (or feasible) to track its development

7
 x
p

Figure 1.2: A constant density ellipse (dashed line) from an initially

bi-Gaussian distribution deforms in the presence of nonlinear forces (solid
line). The smallest ellipse that encloses the final distribution (dotted line) is
clearly larger than the initial ellipse.

through an accelerator. Instead, one is interested in following the beam’s

extent, in x and px , as a function of position along the accelerator. In this

method of analyzing a beam’s behavior, envelope equations are developed

to predict the size of the beam. The advantage of the envelope analysis ap-

proach is that it is a compact description of the beam which easily allows the

inclusion of both space-charge and external forces. This method of analysis is

widely used in the description of photoinjectors, and will be used frequently

in this thesis.

There are two points that need clarification in regard to this approach.

The first is that we are interested in envelopes as a function of position and

not time. The reason for this is simply that for most charged particle beam

systems, the forces involved are given in space, not in time (the position of

8
a focusing element or accelerating structure, for example.) This means that

there is a design path that the beam is to take through the system, and we

are examining the beam at given positions (z in linear systems) along that

path. By definition a beam has the property that its particles have much

smaller transverse momentum than longitudinal (px  pz ≈ ptot , where ptot

2
is the total momentum of the particle, ptot = px2 + py2 + pz2 .) This paraxial

approximation allows us to replace the transverse momentum (px ) with the

slope of the particle trajectory (x
≡ dx/dz) through

dx vx px
= ≈ . (1.14)
dz vz ptot


Similarly, we replace the phase space distribution fx x, px with a trace

space distribution fx (x, x
).

The second issue to address is the definition of the beam “size,” or en-

velope. This definition is easy to make in situations where the distribution

has a “hard edge,” where fx (x, x
) goes to zero rapidly. However, in beams

produced by photoinjectors, as with most beams, this is typically not the

case. In beams like the bi-Gaussian example shown above, there is no hard

edge to the distribution, but rather tails whose extent is determined by the

temperature of the beam. In cases where the loss of beam particles is a con-

cern, such as very high current proton accelerators where dumping the beam

may cause damage or activation of the beamline, it has been customary [5]

to define the beam size such that a given number of beam particles (say 90%)

are inside the envelope. In photoinjectors, this is typically not a concern, and

a method that can be applied consistently to distributions of arbitrary shape

is desirable. This is because, unlike the bi-Gaussian, photoinjector beams do

9
not have self-similar propagation — the shape of the distribution evolves as

the beam moves through the system.

In photoinjector physics, the method commonly used to compare the be-

havior of different distributions uses root-mean-square (rms) envelopes. The

use of rms quantities, based on the second moments of the distribution, was

introduced by Lapostolle and Sacherer [6, 7]. Sacherer showed that the en-

velope equations previously derived for a beam with uniform charge density

applied to any beam with ellipsoidal form, provides that the beam envelope

is defined in the rms sense. This is the concept of equivalent beams, and it

states that if the current (a measure of the space-charge force) and kinetic

energy of two distributions are the same, and the second moments of the

distributions are the same, then they will remain the same when subjected

to the same external forces.

The second moments of a distribution fx (x, x
) are defined as:

  
x 2
= x 2 fx (x, x
) dxdx

  

2
x = x
2 fx (x, x
) dxdx
(1.15)




xx = xx
fx (x, x
) dxdx
.

The spatial rms envelope is then defined as σx ≡ x 2 , and the other rms

quantities are similarly defined as σx
≡ x
2  and σxx
≡ xx
. In the case

of the bi-Gaussian distribution, we saw that an ellipse of constant density

will maintain the same phase space shape and area as long as the applied

and self forces are linear. The same statement can be made of the rms ellipse

governed by the equation

σx2
x 2 − 2σxx
xx
+ σx2 x
2 = εx2 , (1.16)

10
 x

x
σxx

σx
x
= σx2

x
σx

A = π εx

Figure 1.3: Properties of the rms ellipse.

where εx is the rms emittance in the (x, x
) trace plane. The area of the

rms ellipse is π εx . The rms emittance is defined analytically in terms of the

second moments as

εx ≡ x 2  x
2  − xx
2 . (1.17)

The emittance provides a quantitative measure of the beam quality. It de-

scribes the effective trace space area of the beam. This point is illustrated in

Figure 1.2, where the phase space area occupied by a beam in the presence

of nonlinear forces does not change (as dictated by Liouville’s theorem) but

the emittance increases.

The rms emittance as defined by Equation (1.17) is inadequate for the

purpose of comparing beams of different energy because it is defined in terms

of the trace space variables x and x
. The angle with respect to the z-axis
1/2
is defined as px /ptot or px /βγmc, where β = v/c, γ = 1/(1 − β2 ) , m

the electron rest mass, and c the speed of light. As the beam gains energy,

this angle, and the rms emittance, will decrease even though the transverse

11
momentum does not. To address this, the normalized emittance is used,

εx,n = βγεx , (1.18)

where the subscript n is used to specify that the emittance in question is

normalized. It is customary in the accelerator community to quote normal-

ized emittances, and the emittance defined by Equation (1.17) is also known

as the geometric emittance, εx,g .

1.1.3 Emittance and Brightness as Figures of Merit

The product of the two transverse normalized emittances is the standard

figure of merit used to describe the transverse phase space volume of a beam.

To fully specify the quality, however, the number of particles, or current

density, within this volume must be given. In other words, the beam quality

is determined by the density of particles in six-dimensional phase space.

The measure of phase space density is termed brightness, and is given by

definitions that vary based on the beam of interest to a given application.

One such definition gives the normalized brightness as

N
Bn = , (1.19)
εx,n εy,n σz σγ

where σz is the rms bunch length and σγ the rms energy spread. This def-

inition is invariant for linear, time-independent systems [8], although it is

somewhat unsatifactory for photoinjector beams because the rf accelerating

fields are time-dependent. Equation (1.19) ignors correlations in longitudinal

phase space, which can be created and later removed (this is the principle be-

hind magnetic compression), changing the brightness calculated from Equa-

12
tion (1.19), even though the beam’s phase space volume did not change. The

transverse brightness, a definition of the term that is common in describing

the requirements of beam applications, is given by

I
B⊥ = , (1.20)
εx,n εy,n

where I is the average current of the bunch.

The stated goal of producing high quality beams may now be recast in

general terms as the production high brightness — high current, low emit-

tance — beams. The relative importance of these figures of merit depend on

the particular application one is interested in, and in many cases the physical

processes are critically dependent on certain aspects (projections of the 6-D

brightness) such as emittance, current, or energy spread.

One particularly relevant, and demanding application of high brightness

beams is the self-amplified spontaneous emission (SASE) free-electron laser

(FEL) [9, 10]. The SASE-FEL places very specific requirements on the electron

beam parameters, and as a result the brightness is not very useful as a single

figure of merit. Instead, the emittance, energy spread, and beam current all

need to satisfy certain criteria. To begin with, the mean energy of the electron

beam is controlled by the FEL resonance condition for a desired radiation

wavelength λ.

λu K2
λ= 1+ , (1.21)
2γ 2 2
where λu is the undulator period and K is the undulator strength. The “basic”

requirements on the transverse emittance and energy spread are given by FEL

theory [11] as:

γλ
εn ≤ (1.22)
4π

13
 σγ 1
≤ , (1.23)
γ Nu
where Nu is the number of undulator periods. If these requirements are met,

the necessary beam current is derived from the desired gain length.

λu
Lgain = , (1.24)
4π ρ

where ρ is the FEL-parameter [10] defined by

  23
Kλu
ρ= · ωp . (1.25)
8π c

The beam plasma frequency, ωp , contains information about the current I

through
4π re c 2 nb 2c 2 I
ω2p = = , (1.26)
γ3 γ 3 I0 σr2
where re is the classical electron radius, I0 = ec/re is the characteristic cur-

rent and the radial rms width σr2 = σx2 + σy2 is used. The focusing optics in

these systems are typically chosen so that the beam is “matched” — i.e. σr

and σr
are constant and σr r
is zero. In that case, σr2 ∝ εn

2
and the plasma

frequency and therefore the FEL parameter and gain length are determined

by the transverse brightness of the beam. For this reason, the transverse

brightness is a popular figure of merit in FEL applications. Despite this, one

must consider the above equations to determine the importance of the vari-

ous beam parameters on a case by case basis. For example, a situation may

arise where increasing the beam current to the detriment of the emittance

and energy spread may be beneficial to the FEL performance, even though

the brightness may decrease.

These requirements on the emittance, energy spread, and current of the

beam are shown in Table 1.1 for the Linac Coherent Light Source (LCLS) X-ray

14
 LCLS Electron Beam Parameters
Parameter Value Units

Beam Energy 14.35 GeV

Normalized Emittance 1.2 mm mrad

Bunch Length (rms) 20 µm

Peak Current 3.4 kA

Correlated Energy Spread (rms) 0.10 %

Uncorrelated Energy Spread (rms) 0.006 %

Table 1.1: X-ray FEL electron beam requirements

FEL designed to produce coherent radiation with a 1.5Å wavelength [12, 13].

The distinction between correlated and uncorrelated energy spread given in

the table is made because Equations (1.22) and (1.23) do not apply to the

full beam, but rather to individual longitudinal beam slices of length approx-

imately equal to the cooperation length [11]. This distinction is often made

when describing photoinjector beams because both the average energy and

phase space angle of a given slice can vary over the length of the bunch.

The result is that the (projected) rms energy spread and emittance of the full

beam is larger than that of an individual slice, which is initially due to thermal

velocity spread at the beam source. The term slice emittance is typically used

to specify the uncorrelated emittance of a longitudinal beam slice when this

distinction is important — as it is in Table 1.1 because the emittance given

there is not a slice, but a projected emittance.

As one can see, the beam requirements for LCLS are quite a challenge,

15
and have been a motivating force in the development and theoretical under-

standing of photoinjectors. The beam required in the injector (E  100 MeV)

is a dense, short, and cold single-component plasma. As such, its behavior

is dominated by the self-fields it produces. This behavior is fundamentally

different than that of thermally (emittance) dominated beams and the self-

fields can cause deterioration of the beam brightness during acceleration,

transport, and subsequent handling (such as pulse compression) of the beam.

The physical phenomena involved in these processes form the central theme

of this thesis.

To begin this survey of topics in high brightness beam physics, the re-

mainder of this chapter examines how the photoinjector improves on the

creation of high-brightness beams over previous sources, and discusses the

Neptune photoinjector in terms of the performance required of its compo-

nents. Chapter 2 is focused on the evolution of the transverse emittance in

accelerated and transported space-charge dominated, photoinjector beams.

Chapter 3 examines the effect space-charge fields have on the measurement

of these low-emittance beams. Finally, in Chapter 4 the process of magnetic

pulse compression is investigated, and experimental evidence of transverse

phase space distortions due to space-charge fields is presented and analyzed.

1.2 Photoinjector Development

The key technological developments that led to the photoinjector were the

high gradient rf accelerating cavity and the photo-cathode, as illuminated

by an ultra-short pulse laser. The importance of these advances can be un-

16
derstood by examining the limitations of previous sources and how the new

technologies overcome those difficulties.

1.2.1 DC Guns and Thermionic Cathodes

The thermionic DC gun is used in beam-related applications such as klystrons,

for high power microwave generation, sources for accelerators, and more

recently, have been proposed for use in electron beam cleaning of photo-

cathodes [14]. This type of source is illustrated in Figure 1.4, and consists of

a pair of electrode plates with a voltage drop V between them. The cathode

(when dealing with electrons) contains a thermionic emitter that produces

electrons when heated. The electrons are accelerated through the gap be-

tween the cathode and anode by the voltage drop across the two electrodes

and exits the gun through a hole in the anode. In practice, the cathode and

anode are usually not planar, but are curved, so that the electric field be-

tween the two provides some radial focusing of the emitted beam in order

to counteract radial space-charge induced expansion. Guns that employ this

focusing geometry are typically referred to as Pierce-type [15] guns.

In this type of gun, the size of the beam is determined by the size of

the emitting surface of the cathode, which is usually circular with radius

rc . The conduction electrons in the emitter behave as a Fermi-Dirac gas. If

the cathode temperature is zero, these free electrons will populate all of the

available energy states up to the Fermi energy EF , and no states above EF .

For a finite temperature, the probability for an energy state to be occupied

transitions from one to zero in a tail centered at the Fermi energy, with a

17
 : E q u pi o et ntial (I= 50 m A)
60
B eam En ergy = 3 keV

40
R a d iu s

20
r=4 .5 m m

0
0 20 40 60 80 100 120 1 40
d =1 .1 cm
A x ia l D i stan ce

Figure 1.4: r versus z schematic of an axially symmetric Pierce type DC

gun with thermionic cathode. This gun was designed to do electron beam
cleaning of a photo-cathode [14]. The design was copied from guns used
at the Maryland storage ring [16]. Here the equipotential lines shown are
shaped by the cathode and anode surfaces to focus the beam emitted from
the thermionic source, as shown by the lines defining particle trajectories.

width characterized by the temperature. The electrons in this tail emerge

from the cathode having an uncorrelated velocity spread controlled by the

temperature of the emitter, given by


kB T
σvx = (1.27)
m

where kB is Boltzman’s constant. If the beam density is constant, then its

rms width is σx = rc /2, and the intrinsic emittance of the source is given by

σvx rc kB T
εn = σx = . (1.28)
c 2 mc 2

The amount of current produced may be limited either by the temperature

of the emitter, or by the longitudinal space-charge forces in the beam. In

18
klystrons, space-charge is the limiting factor. Here, the accelerating electric

field in the gun is partially canceled by the field of opposite direction, set

up between the charge of emitted particles and the cathode. As the beam

current increases, this cancellation increases, until the fields are equal, and

the current cannot be increased further. This balance point is specified by

the Child-Langmuir law [17],

I = KV 3/2 , (1.29)

where K is the beam perveance, and is determined by the geometry of the

gun. In this limit, then the beam current can be increased by raising the

voltage between the electrodes. The maximum voltage difference one can

generate is limited by high-voltage breakdown to about 500 keV [3]. In appli-

cations such as klystrons or injectors for accelerators, the beam is required

to be bunched after the gun. In accelerators, bunched beams are needed to

minimize energy spread in the rf cavities used to accelerate the beam, as

discussed below. Bunchers and pre-bunchers are rf cavities that bunch the

beam by producing a sinusoidally time varying accelerating field. Near the

zero crossing of the field, the momentum gain (or loss) of particles is linear,

with the higher momentum particles in the rear. When these particles are

allowed to drift after the bunching cavity, the longitudinal velocity modula-

tion of the beam becomes a spatial modulation — the beam bunches. This

velocity bunching process becomes much less effective when the beam is rel-

ativistic as the velocity of a particle is no longer linearly dependent on its

momentum. This fact sets a practical limit on the desired output energy of

DC guns which, for electrons, is roughly 400 keV.

Thermionic guns used as injectors for accelerators are typically run at

19
high voltages in order to minimize the effects of transverse space-charge

forces. At these voltage levels, the current density required for the beam

to be Child-Langmuir limited is higher than the thermal limit of the emitter.

In this limit the current density at the cathode is given by the Richardson-

Dushman equation [18, 19],

J = AT 2 e−W /kB T , (1.30)

where W is the work function of the emitter material and A is a constant

derived from fundamental physical constants. If we again assume a con-


stant current density across the emitting surface, (so rc = I/π J) then Equa-

tion (1.28) can be rewritten as

 
1 I kB T
εn = · . (1.31)
2 πJ mc 2

Using this and Equation (1.20) the transverse brightness for this source is

mc 2
B⊥ = 4π J . (1.32)
kB T

This equation gives the maximum brightness achievable at a thermionic source.

In a slightly modified form it applies to photo-cathode produced beams as

well, and it therefore serves to emphasize the important properties of differ-

ent sources (namely, the current density and temperature.) For a dispenser

type cathode, typical operating conditions give J from 10 to 20 Amps/cm2

and kB T about 0.1-0.2 eV [3]. Applying these parameters to Equation (1.32)

gives an upper bound to the transverse brightness of thermionic sources of

1013 Amps/m2 . The brightness required by LCLS computed from Table 1.1

is two orders of magnitude higher than this. The comparison here is not en-

tirely fair, since the current of the beam can be increased by buncher cavities

20
and magnetic compressors at relativistic energies (but the emittance and en-

ergy spread will increase as well!) Nonetheless, it was realized in the 1980s

that the current densities and therefore brightness of thermionic guns were

not high enough for SASE-FEL and other high-brightness beam applications

[20].

1.2.2 RF Cavities and Photo-cathodes

Radio frequency cavities are capable of generating much higher accelerating

fields than DC guns. The achievable peak field is an increasing function of the

resonant frequency of the cavity, and at S-band (f = 2856 MHz) 140 MV/m

electric fields have been produced [21]. The ability to quickly accelerate a

high current beam to relativistic energies is a great advantage in mitigating

the effects of transverse space-charge forces on the beam. This is because, as

will be shown in Chapter 2, the radial defocusing force due to space-charge

fields varies as 1/γ 2 . RF guns with thermionic cathodes were developed in

order to utilize this property [22]. This type of gun is illustrated in Figure 1.5.

The time-dependent nature of the rf accelerating fields has other, less

desirable effects on the beam. Because the availability of electron from the

thermionic cathode is constant over an rf period — the current increases with

accelerating cathode field up to the thermal limit imposed by Equation (1.30)

— different parts of the beam are exposed to the full range of accelerating

and focusing field strengths in the rf cavities. The result is that the beam that

emerges has a continuous energy distribution from zero to the maximum de-

termined by the strength of the rf fields. Likewise, beam slices at different rf

21
 r
cathode

Figure 1.5: A cylindrically symmetric, standing wave, single cell rf gun. The
magnitude of the fields vary as sin(ωrf t).

phases are exposed to different focusing fields and therefore had differing

sizes and phase space correlations, resulting in an increased emittance [23].

One method used to counter this is to bend the beam with a dipole mag-

net. This separates slices of different energy transversely, and allows one to

collimate the beam and use only a selected portion.

The extremely high accelerating fields and collimation of the highest qual-

ity portion of the beam serve to minimize the emittance growth of the therm-

ionic beam. They do not change the intrinsic emittance of the source given

by Equation (1.31). Photo-cathodes address this issue. This source utilizes

the photo-electric effect to produce electrons. In photoemission, the cathode

is illuminated with a laser pulse, where the photon energy
ω is greater than

the work function W of the metal cathode (semi-conductors are also used

as photo-cathodes, and there the work function is replaced by the band gap

energy so
ω > EBG .) Electrons emitted from a photo-cathode have an uncor-

22
related transverse velocity spread just as in the case of thermionic cathodes,

although the photoemission process is more complicated, and the velocity

spread is not a function of temperature alone [24]. Despite this an effective

source temperature Te can be defined such that Equation (1.28) holds for

photo-beams, where T is replaced by Te . Theoretical considerations [24] and

experimental evidence [25] suggest that the effective temperature for condi-

tions typical of photoinjectors is around kB Te from 0.1 to 0.2 eV. This leads to

a thermal emittance of approximately 0.3 mm mrad per mm cathode radius,

which is a generally accepted figure [26, 20].

It is interesting to note that the effective temperature of thermionic and

photo emitters is basically the same. The real decrease in intrinsic emittance

and increase in brightness for photo-beams comes from the higher current

density they produce. Since the electrons in a photo-cathode are directly pro-

moted by absorption of a photon to an energy state above the work function,

the number of electrons available for acceleration depends on the number of

photon hitting the cathode, and not on a thermalization process in the Fermi

gas. The result is current densities at the level of a few thousand Amps/cm2

[27]. From Equation (1.32) we see that the intrinsic transverse brightness

of photo-cathode beams is two orders of magnitude higher than that of a

thermionic source. There are further advantages to using photo-cathodes

with rf guns — the device known as the photoinjector.

23
 Beam out

Laser in
photo-
cathode

Figure 1.6: A possible photo-cathode rf gun.

1.2.3 Photoinjectors

The first photoinjector was built and operated in the mid-1980s at Los Alamos

National Lab [28]. Photoinjectors combine the beneficial aspects of photo-

cathodes and rf guns (technically, a photo-cathode in a DC gun is also a

photoinjector, but the vast majority of photoinjectors use rf guns, and the

term photoinjector is generally taken to mean rf photoinjector.) As an exam-

ple, consider the illustration of a possible photoinjector gun in Figure 1.6. In

this example, the gun consists of several rf cavities or “cells” electromagneti-

cally coupled by the axial openings between cells that allow the beam to pass

through the structure. The field configuration indicated in the schematic is

that of a standing-wave π -mode. On axis, it has the form [29]

∞

Ez (z, t) = E0 an cos (nkz) sin (ωt + φ0 ) , (1.33)
n=1

where φ0 is the rf phase at the cathode z = 0 at time t = 0, k is the rf

wave number and ω = ck. This equation can be used as a starting point to

calculate the motion of single particles in the gun [30]. From this type of

24
analysis, Kim showed [23] that the time-dependent rf forces in the gun re-

sulted in emittance and energy spread increases that depend on the dimen-

sions of the beam — laser pulse — on the cathode. Specifically, he showed

that σγ ∝ (kσz )2 and εx,rf ∝ k3 σx2 σz2 , where σγ,x,z refer to the rms energy

spread, and transverse and longitudinal dimensions of the laser pulse, re-

spectively. It is advantageous therefore, to uses laser pulses that are much

smaller than the rf wavelength λrf = 2π /k  σx,z . At S-band, λrf = 10.5 cm,

and the laser dimensions used are typically on the order of one mm. This

automatic production of bunched beams eliminates the need to collimate

the beam as in the case of thermionic rf guns, and results in small energy

spreads and rf-induced emittance. Typical numbers for millimeter laser sizes

at S-band are <0.1% energy spread and 0.1-0.3 mm mrad rf-induced emittance

[31]. The small beam size, combined with the extreme current density of the

photo-cathode source result in very strong self-fields. These fields are so

strong – and the thermal and rf emittance is so small — that the behavior of

photoinjector beams is determined almost exclusively by space-charge. The

emittance and energy spread induced by these space-charge fields can be

quite large if not adequately compensated, and this compensation is critical

in maintaining the advantage photoinjectors have in high brightness beam

production. The theoretical models in Chapter 2 address this issue.

1.3 The Neptune Photoinjector

The use of photoinjectors to produce high-brightness electron beams can

place demanding technical and operational requirements on the various com-

25
ponents that make up the system. As a more thorough introduction to pho-

toinjectors, the photoinjector in the Neptune Advanced Accelerator Labora-

tory at UCLA is described below. The limitations and required performance

of the major photoinjector components are discussed with regard to the pro-

duction of the high brightness beams required by the laboratory.

1.3.1 The Neptune Advanced Accelerator Laboratory

The Neptune Advanced Accelerator Laboratory was constructed to perform a

variety of laser-plasma and beam-plasma advanced acceleration experiments

[32]. The primary mission of the lab is to accelerate a high phase space den-

sity beam in a Plasma Beat-Wave Accelerator (PBWA) to an energy of 100

MeV while maintaining the beam brightness [33]. To that end, the lab con-

sists of two major pieces: the photoinjector [34] and the two wavelength,

100 picosecond, one terawatt MARS [35] CO2 laser. A schematic layout of

the laboratory is shown in Figure 1.7. In addition to the PBWA, experiments

in plasma wake-field acceleration [36], inverse FEL acceleration [37], plasma

lensing [38], and other advanced accelerator experiments are planned. The

photoinjector itself provides a means to do experiments in fundamental high-

brightness beam physics, and one such experiment is described in Chapter 4

of this thesis. Like the LCLS, the PBWA is an example of an application that re-

quires high brightness beams. For the MARS laser, the beat-wavelength is 330

µm. This determines the dimensions of the accelerating “cavity,” and implies

that the beam dimensions — both transverse and longitudinal — must be sig-

nificantly smaller than 330 µm in order to preserve the quality of the beam

26
 Photoinjector
Diagnostic
Area

Bunker

1.5
.5 R
5 m RR

MARS
Laser
Photo-cathode
drive laser
transport line

Control
Laser Room
Room

Photo-cathode
drive laser

Figure 1.7: The Neptune Advanced Accelerator Laboratory

27
[33]. Small transverse spot sizes require low emittance and energy spread,

while extremely short bunches require compression of the photo-beam [39].

In order to produce the type of beam required by the PBWA, and other exper-

iments, the injector must be stable, flexible, and well diagnosed. These are

typical photoinjector design considerations, and as such, the implemented

hardware reflects these features.

The major components of the Neptune photoinjector are the accelerating

structures, photo-cathode drive laser system, rf system, bunch compressor,

beam diagnostics, and control system. The beamline itself — consisting of

vacuum hardware and magnetic optics in addition to the accelerators, com-

pressor and diagnostics — is shown in Figure 1.8.

1.3.2 Accelerating Sections

The beam in the Neptune photoinjector is accelerated using a split system

composed of a high gradient rf photo-cathode gun, followed by a drift length

and then a moderate gradient booster linac. The gun is a 1.625 cell, S-band, π -

mode, standing-wave structure produced by a BNL-SLAC-UCLA collaboration

[21]. This gun is the result of several generations of design improvements

that result in high quality and high strength accelerating fields. Rf is fed

into the gun through a coupling slot in the full cell cavity, and coupled into

the half cell on-axis through the iris. The rf slot is symmetrized with an

identical opening on the opposite side of the full cell. In fact, the other off-

axis openings in the cells — the 70 degree laser injection ports in the half

cell, and tuning ports in the full cell — are symmetrized in order to minimize

28
 5' x 3' optical table

Cu photo-cathode
1.625 cell rf gun

Emittance compensating
solenoids
Laser mirror box

Steering magnets
YAG screen

7+2/2 cell PWT linac

YAG screen

Steering magnets Quadrupole tripplet

Phosphor screen
Integrating current
transformer

Chicane compressor/
energy specrometer

Phosphor screen

Emittance slits

Quadrupole tripplet
CTR foil

YAG screen
Steering magnets

Figure 1.8: The Neptune photoinjector beamline.

29
dipole components to the fields that can lead to emittance growth [40]. The

gun is normally operated with a peak accelerating field of Ez = 85 MV/m,

and the beam energy exiting the gun is roughly 4 MeV.

The back-plane of the half cell is a removable piece that contains the cath-

ode at its center. The relative ease with which the cathode-plane can be re-

moved allows the quick replacement of cathodes, and several have been used

thus far. Currently, the cathode consists of a single crystal of copper brazed

into the OFHC copper back-plane. Copper was chosen as the cathode mate-

rial because of its relatively easy vacuum requirements, and its long lifetime

(both vacuum and lifetime considerations are much more stringent for ma-

terials such as semi-conductors [41]). The disadvantage of using copper is

its comparatively low quantum efficiency (QE) — defined as the number of

electrons produced per incident photon — but the amount of charge required

for the planned PBWA experiments is less than one nano-coulomb, and the

QE of copper combined with the available laser energy is sufficient to pro-

duce this charge. The single crystal cathode has the potential to emit charge

more uniformly than amorphous copper, and data comparing the emission

of single-crystal and amorphous copper cathodes indicates that the single

crystal cathode has a factor of 2 higher QE [42].

The 4 MeV beam exiting the gun is focused by a solenoid magnet and al-

lowed to drift approximately one meter to the entrance of the booster linac.

This focusing is critical to the space-charge dynamics and emittance evolu-

tion of the beam in both the drift section and the following linac, and will be

described in detail in Chapter 2. The linac is a 7 + 2/2 cell, π -mode, standing-

wave structure. Its design is that of a plane-wave transformer (PWT) [43]. The

30
PWT differs from the standard disk-loaded wave guide structure depicted in

Figure 1.6 in that the disks do not extend to the outer wall of the tank, but

rather stop short, leaving a coaxial gap. This gap functions as a plane-wave

transmission line, and as a result, the structure has extremely strong cell-

to-cell coupling and robust mode separation. The large mode separation

permits the use of a large number of standing-wave cells. The PWT was de-

signed to produce a peak accelerating field of Ez = 60 MV/m but is normally

run at a slightly lower gradient around 50 MV/m. The fully accelerated beam

exiting the PWT in the Neptune system has a nominal energy of 12 MeV.

With respect to beam quality, the important properties of the accelera-

tors are the quality and consistency of the electromagnetic fields; quality

measured in terms of field balance, transverse multipole moments, and longi-

tudinal field shape (spatial harmonics) considerations, and consistency mea-

sured in terms of the stability of rf amplitude and phase, both shot-to-shot

and on a time scale needed to preform experiments. Field shape and symme-

try are issues considered mostly in the design of the structures, and in both

the gun [40] and the PWT [find Rhen thesis] efforts were made to symmetrize

the cavities as much as possible. Field balance — the ratio of Ez measured

in the center of adjacent cells — is accomplished during the final tuning pro-

cess. The relative cell-to-cell field strengths are measured directly with the

bead-pull technique [44] and indirectly, in the case of the gun, by measur-

ing the frequency separation of the 0-mode and π -mode resonances [45].

An example of the mode separation measurement is shown in Figure 1.9 for

the 1.6 cell gun, taken before installation into the Neptune beamline. Using

bead pulls and mode separation measurements, the fields in the Neptune

31
 5

S11 [dB]
-5
∆ƒ= 3.4 MHz

-10

π-mode
-15

-20 0-mode
2851 2852 2853 2854 2855 2856 2857
Frequency [MHz]

Figure 1.9: Network analyzer measurement of the resonant modes of the

1.6 cell rf gun at Neptune. The frequency separation of the 0 and π modes
indicates the relative strength of the fields in the half and full cells.

gun have been balanced to within a few percent. For the PWT, the strong

cell-to-cell coupling assures good field balance, as confirmed by bead pull

measurements.

Phase and amplitude stability of the rf fields in the cavities is largely gov-

erned by the stability of the electromagnetic power supplied by the rf system.

This is certainly true of shot-to-shot fluctuations, assuming the accelerators

are not suffering from high-voltage breakdown. Longer time scale drifting in

phase or power depends as well, on the rf system, but in addition depends on

the temperature stability of the cavities. For S-band structures, the engineer-

ing rule of thumb is that the resonant frequency of a given mode shifts by

roughly 40 kHz/◦C [21]. This rule implies that the temperature of the accel-

erators must be maintained at the level of 0.1◦ C in order to maintain correct

32
resonant conditions. Both the gun and the PWT are designed with water cool-

ing channels to stabilize the temperature. For the gun, water cooling works

very well, as the amount of power deposited by the rf in the structure is much

lower than the capacity of the water bath to remove it. Unfortunately, the

PWT does not fair as well. Because the PWT structure has the cell-separating

irises removed from the outer wall of the tank, support rods are required

to hold the irises in place. The support rods also function as water cooling

channels and allow for temperature control of the irises. The outer wall of the

structure, however has no direct water cooling. The result of this, is that as rf

power is deposited in the PWT, the inner and outer parts heat differently, and

maintaining the tune of the device becomes difficult. A number of different

methods of solution to this problem have been tried with varying degrees of

success. One of the more successful tactics has been to independently heat

and cool the outer wall of the PWT tank with a fan and electric heater tape,

while maintaining a constant temperature in the interior with the water bath.

This method has proven sufficient to keep the phase from drifting by more

that a few degrees of rf over about one hour; the time required to preform

the experiments described in Chapter 4. The PBWA experiments require this

level of stability from the PWT for several hours at a time, and more robust

temperature control is being developed to provide the required stability. On

any time scale longer that shot-to-shot, the measurement of rf phase in the

PWT is required to correct for drifts due to temperature variation or other

sources of phase drift. The measurement of rf phases is considered part of

the rf system, and will be described below.

33
1.3.3 Photo-cathode Drive Laser

A photoinjector places a number of demands on its drive laser. Some are

specific and strictly required for the photoinjector to operate, while others

apply to differing degrees and affect the quality of the electron beam pro-

duced. Some of the requirements are:

• The wavelength must be short enough for photo-emission (266 nm for

copper cathodes).

• The laser energy must be high enough to produce the desired charge

with a given cathode QE.

• The laser pulse length should be a few tens of rf degrees or less (at

S-band 1◦ ≈ 1 pico-second).

 Minimization of shot-to-shot fluctuations and long-term drift in energy.

 Minimization of fluctuations and drift in laser position on the photo-

cathode.

 Transverse and longitudinal pulse shape should be controlled, and as

uniform as possible.

In the above list, bulleted (•) items are required, while stared () items affect

electron beam quality.

The hardware used to generate photo-electrons in the gun is a chirped-

pulse-amplification (CPA) system [34]. The laser system, shown in Figure 1.10

begins with a 1064 nm mode-locked Nd:YAG laser. The laser is mode-locked

34
 532 nm,
To Doubling Crystal 1 mJ/pulse,
in Bunker 3 ps

2xω
Autocorrelator
BBO Crystal

1/4 Meter
Grating Pair
Spectrometer

2.4 mJ/pulse,
< 2 ps 0.5 km Fiber

Coherent Antares
YAG Oscillator Continuum Nd:Glass
Regenerative Amplifier
Fast
1.064 µm, 76 MHz, Photodiode 5 Hz, 4 mJ/pulse,
4 nJ/pulse, 80 ps Amplification > 10
6

Figure 1.10: The Neptune photoinjector drive laser.

with a 38.08 MHz crystal oscillator. The signal from this crystal is also split

off and frequency multiplied by 75 to produce the 2856 MHz signal that

begins the rf system. In this way, laser pulses are synchronized to the rf.

The laser pulses are sent through a half kilometer long fiber where they are

stretched and “chirped,” that is, a correlation is made between frequency and

position. The chirped pulse exiting the fiber is suitable for amplification in

the Nd:glass regenerative amplifier (regen). The regen accepts laser pulses

at a rate of 5 Hz through a Pockel’s cell that is triggered by the union of

a rough timing 5 Hz trigger and the 38.08 MHz oscillator. The 5 Hz laser

pulses are amplified in the regen by a factor of 106 and then sent to a grat-

ing pair where they are compressed by removing the frequency-chirp. The

pulse length after compression is determined by the separation between the

35
gratings. The minimum laser pulse length measured in this system is 1 ps

(rms), while the gratings are typically set to produce 4 ps (rms) laser pulses.

At this point the laser pulses have the appropriate length for use in the pho-

toinjector, and several mJ of energy per pulse, but the wavelength is a factor

of 4 too long for photo-emission. The laser system was designed to make use

of nonlinear, frequency-doubling crystals to convert the 1064 nm (infrared)

light to 532 nm (green), and again to 266 nm (ultra-violet) at the expense of

energy. The results of two stages of frequency doubling are 100 µJ, ultra-

violet pulses that satisfy the first three requirements of the photoinjector,

producing approximately 1 nC of charge per pulse on a copper cathode.

Of the starred items in the above list of drive laser attributes, the issue

of beam position stability is the most seriously addressed at Neptune (and is

arguably the most important, since it affects the other two.) The green light

produced by the doubling crystal in the laser room is propagated through

a transport system into the bunker, where the second crystal doubles again

to produce the ultra-violet (UV) light sent to the photo-cathode. There are

two factors here that make the laser position important. The first is that

the doubling crystal’s efficiency and output pulse shape depend on both the

input position and angle of the green light [46]. The second is that the elec-

tron pulse should originate in the electrical center of the cathode in order to

avoid steering errors in the system which can lead to emittance growth [3].

Therefore, if the beam drifts away from the optimal position on the UV dou-

bling crystal, the energy and quality of the UV pulse will be diminished, and

it will hit the cathode off-center. The laser transport system compensates

for this potential problem by using a feed-back system to control the laser

36
position and angle. In this system a “beacon” GreNe laser (λ = 543 nm) is co-

propagated with the green laser pulses in the transport system. The beacon

laser is inserted and extracted from the transport line using beam splitters

that either reflect or transmit light based on its polarization. Once the beacon

and drive lasers are aligned, the drive laser is positioned correctly on the UV

doubling crystal. At this point, the position and angle of the split-off beacon

laser is measured by a pair of 4-segment photo-diodes (quadrant detectors).

These photo-diodes send signals to a computer, which monitors the position

of the laser. The transport system has a pair of motorized mirrors that the

computer controls to periodically re-center the beacon signal on the photo-

diodes. The feed-back system effectively eliminates drifts in laser position

that are on the time scale of one minute or longer. Shot-to-shot fluctuations

are not handled, and these have been measured to be less that 10% (rms) [47]

of the laser spot size on the cathode.

While the feed-back system takes care not to exacerbate problems with

laser energy fluctuations, it does not reduce them. Unfortunately, the many

nonlinear effects in the laser system make energy fluctuations rather large

and shot-to-shot energy jitter is at best 10% (rms.) This jitter, combined with

gradual drift in energy, make electron beam charge an important diagnos-

tic in the operation of a photoinjector. Thus single-shot, non-destructive

measurements of beam parameters as a function of charge are particularly

valuable.

The shape of the laser pulse (at least initially) determines the shape of

the electron beam emitted. As will be shown in Chapter 2, both the trans-

verse and longitudinal beam profiles effect the behavior, and emittance of the

37
 38.08 MHz
phase Klystron
oscillator shifter high power
2856 MHz SF6-filled
coaxial cable waveguide
kW
f x 75 Φ amp
XK5
18 MW
(low level rf) 12:1 pulse
transformer
high power
Laser oscillator circulator
(mode-locker) Thyratron
4 µsec
modulator
12 MW
timing 4.77 dB
stabilizer Φ Attenuator
splitter
6 MW

50 kV constant 1.6 cell Attenuator

current supply PWT gun

Figure 1.11: The Neptune photoinjector rf system.

final beam, with uniform, hard-edge distributions favorable to smooth dis-

tributions with tails. Because of this, several photoinjector laboratories have

used different techniques to shape the laser pulses [48, 49, 50]. Currently at

Neptune no pulse shaping is done. The profile of the UV is approximately

Gaussian both transversely and longitudinally.

1.3.4 RF System

The RF system, as shown in Figure 1.11, begins with a mW level 2856 MHz

signal derived from the crystal oscillator used to mode-lock the laser, as

described above. The low-level rf is then transported to the control room,

where a manual phase shifter allows the operator to control the overall phase

38
between the laser and the rf. This control is used to set the proper laser

injection phase in the gun, and to verify that the laser position on the cathode

is correct (if the beam is off-center, the radial focusing and defocusing fields

in the gun will steer the beam to one side, allowing one to test for centering

by sweeping through rf phase and observing motion of the electron beam

centroid downstream of the gun.) The low-level signal is then sent to a pair

of pre-amplifiers and the output one Watt signal is routed to the input of the

kiloWatt amplifier. The kiloWatt amplifier (kW-amp) is a pulsed, traveling

wave tube (TWT). The kW-amp outputs a 10 µsec, 600 W pulse that is fed via

a high power rf cable to the input of a SLAC XK-5 klystron.

The klystron is a classic example of a DC thermionic gun discussed in

Section 1.2.1. The klystron operates in the Child-Langmuir limit and the

output electron beam is bunched by the feed rf. The bunched electron beam

then interacts with a series of coupled rf cavities resulting in an instability

that converts a significant percentage of the beam power into rf power [51].

The beam voltage in the klystron is produced by a roughly flat, 4 µsec, high

voltage pulse, which is timed to arrive near the end of the input rf pulse. The

high voltage pulse is generated by charging, to roughly 40 kV, a capacitor

bank that is part of the modulator’s pulse forming network (PFN). The PFN is

made of a series of capacitors and inductors that act as a transmission line.

The PFN is fired by a thyratron switch, producing a 20 kV pulse impedance

matched to the klystron. The pulse is sent through a 12:1 DC-core-biased

transformer and drives the klystron cathode.

The klystron can produce up to 25 MW of rf power, but is typically run

at about 18 MW, limited by rf breakdown in the accelerators and wave guide.

39
The wave guide system distributes power to the gun and PWT. Because both

accelerating structures operate in standing wave modes, they reflect power

at the beginning and end of the rf pulse. To protect the klystron from this

reflected power, and reflections due to breakdowns, the first element in the

wave guide system is a 4-port circulator that acts as an isolator. After the

isolator, the rf is split by a 4.77 dB divider, sending 2/3 to the PWT and the

rest to the gun. In the two branches of wave guide is a motorized attenuator,

which allows independent control of the power delivered to each of the struc-

tures. On the PWT branch there is a high power, motorized phase shifter as

well. This permits the operator to adjust the phase of the PWT with respect

to the gun, for proper injection of the electron beam that achieves optimum

acceleration.

As was discussed in Section 1.3.2, the phase of the rf waves in the accel-

erators is an important parameter in operating a photoinjector. In addition

to tuning of the cavities and laser stability, the rf system is certainly a source

of potential phase fluctuations and drifts. The active elements of the system,

namely the kW-amp or high voltage inputs to the klystron, or the klystron

itself, are the most likely sources. Regardless of source, the fact that phase

jitters and drifts affect the properties of the electron beam makes the direct

measurement and control of the cavity phases necessary. In the Neptune sys-

tem, the phase of the gun and PWT are both measured with respect to low-

level rf using quadrature mixers. By using both outputs of the quadrature

mixer [one proportional to sin (φ), and the other to cos (φ)] single-shot mea-

surements independent of rf power can be made, as illustrated in Figure 1.12.

This measurement method allows manual correction of the phases over time,

40
 150 60

)
n (φ
100 40

A si
50 20
Mixer Signal [mV]

PWT Phase [◦ ]
)
φ
0 0

s(
co
A
-50 -20

-100 -40
phase

-150 -60
-5 0 5 10 15 20

Time [µsec]

Figure 1.12: The PWT phase is measured using the outputs of the quadrature
mixer. The plot shows oscilloscope traces of the mixer signals and the phase
over one rf pulse.

which was done in the experiment described in Chapter 4. A feed-back sys-

tem can be implemented here in a similar fashion to the laser feed-back, and

this work is currently in progress.

1.3.5 Magnetic Compressor

As a major component of the Neptune photoinjector, the magnetic bunch

compressor deserves a brief description at this point, particularly since this

type of device is either in use, or under design in many photoinjector labs

41
 2' 4"

Figure 1.13: The Neptune chicane; it can be used in compressor or spec-

trometer mode.

[39]. The use of a compressor of this type can dramatically change the elec-

tron beam in both longitudinal and transverse phase space due to collective

effects. The physical processes that cause this change, space-charge and

radiation effects, are investigated in Chapter 4. A brief description of the

compressor hardware is given here to illustrate its operating principle and

indicate the precision required in its design and implementation.

The Neptune compressor, shown in Figure 1.13, is made of four dipole

magnets configured with the field in the first and fourth magnets in the op-

posite direction to that in the middle two. In this “chicane” configuration,

as the figure shows, beam particles with higher momentum take a shorter

42
path than those with lower momentum. If the short drift spaces between the

dipoles are ignored, the total path length of a particle with a given momentum

p is just 4 times its path in one magnet, or



S = 4 × Rbend p θbend p
 
p eBl
= 4 sin−1 , (1.34)
eB p
where Rbend and θbend are the radius of curvature and bend angle in one

magnet, B is the magnetic field, e is the electron charge, and l is the length

of one magnet, 12 cm in this case. The variation in path length with particle

momentum is found by differentiating Equation (1.34):

 
δS tan (θbend ) δp
= 1− (1.35)
S θbend p
δp
≡ αc ,
p

δp
where αc denotes the path length parameter, δS
S p
. When a design bend an-

gle is set between 0 and π /2, a positive δp/p gives a negative δS/S, verifying

the statement that higher momentum particles take a shorter path.

If the electron bunch is set up with a correlation between momentum

and longitudinal position, such that the higher momentum particles trail the

lower momentum particles, this path length difference can be used to com-

press the bunch. The final position of a particle within the bunch depends not

only on its path length difference in the compressor, but also the velocity dif-




ference; that is, it depends on the particle’s time of flight, t p = S p /v p .

This leads to
δt δS δv
= −
t S v 
1 δp
 αc − 2 . (1.36)
γ p

43
In accelerator jargon, the bracketed term is called the momentum compaction,

ηt . The momentum compaction — also known as the time dispersion — is

a useful quantity in periodic systems. In single pass devices like this com-

pressor; however, the problem is usually formulated in analogy with trans-

verse trace space variables. In longitudinal trace space the spatial coordinate,

ζ = −v0 (t − t0 ) is found with respect to the centroid of the bunch, and the

“angle” is the normalized momentum error ζ
= δpz /p  δp/p. In analogy

with transverse focusing, the compressor acts like a drift length, allowing the

trace space correlation to remove itself and compress (focus) the beam. In

transport matrix notation, the longitudinal drift is given by R56 ,

δζ
R56 = = −ηt δS. (1.37)
δp/p

The Neptune chicane is designed with a bend radius of 31 cm and a bend

angle of 22.5 degrees. At the nominal beam energy of 12 MeV, this leads to

an R56 of approximately 1 cm (the computation of R56 is complicated slightly

when the drifts between the magnets are included, but the procedure is the

same).

The above formalism shows how a longitudinal phase space correlation

is removed with the chicane. The correlation must, of course, first be intro-

duced, and this is done with the PWT. The violent acceleration of the beam in

the PWT dominates its longitudinal dynamics, and as a result, the trace space

correlation is determined by the accelerating field at the injection phase, φ0 .

The momentum distribution across the bunch is approximately given by:

pz (ζ)  pmax cos (krf ζ + φ0 ) , (1.38)

where pmax is highest possible momentum achieved by injecting “on crest,”

44
or at φ0 = 0, and krf = 2π /λrf is the rf wave number. Using a trigonometric

addition formula and expanding to second order in krf ζ gives

 
1 2
pz (ζ)  p0 1 − (krf ζ) − tan (φ0 ) krf ζ , (1.39)
2
where p0 = pmax cos (φ0 ) is the momentum of the bunch centroid.

It can be seen from this that the linear correlation between the longi-

tudinal trace space variables is − tan (φ0 ) krf . It is also clear from this ex-

pression that the momentum distribution contains a nonlinear component

that depends on the rf phase extent of the beam. This rf-induced non-

linear correlation is the most significant source of longitudinal emittance,

   2
(εζ ≡ ζ 2 ζ
2 − ζζ
) and in keeping with the transverse dynamics

analogy, the emittance controls the minimum beam size after a lens (PWT

injection phase) and drift length (chicane).

In any case, Equation (1.39) can be differentiated to get an expression

comparable to the R56 of the compressor,

δζ 1 1
=− · . (1.40)
δp/p krf tan (φ0 ) + krf ζ
Thus, given a desired beam energy, rf wavelength, and initial electron bunch

length, the compression system is designed by choosing the PWT injection

phase and dipole bend angle so that the right hand sides of Equations (1.37)

and (1.40) are equal. The choice is a balancing act between running too far

ahead of crest in the PWT, losing beam energy and increasing momentum

spread, and having too large a bend angle, effecting the transverse focusing

of the beam and self-field processes that lead to emittance growth. For the

Neptune compressor, the compact design requires a fairly large (22.5◦ ) bend

angle to use an injection phase of approximately 25◦ ahead of crest.

45
 The compressor as depicted in Figure 1.13 focuses in both transverse

directions. The focusing it the bend plane (x) is intentionally added to the

system with the edge angles on the entrance of the first, and exit of the last

magnets. Without these edge angles, the chicane has vertical (y) focusing

at each of the magnet edges that the beam passes at the (large) bend angle.

This results in an effective focal length that causes a ballistic beam waist in

the middle of the last dipole. In x, however, path-length focusing from the

entrance and exit angles of the first and second magnets and third and fourth

magnets cancel each other giving zero net focusing. The 11.3◦ entrance and

exit angles of the chicane ameliorate both problems by providing defocusing

in y and focusing in x.

From an operational point of view, use of the compressor requires stability

in the rf injection phase and precision alignment of the magnets and control

of the fields. The common thread of phase stability has been discussed in

relation to every photoinjector component so far, and effects compression as

well. As Equation (1.40) indicates, and pulse length measurements in Chapter

4 will show, the amount of compression and, therefore, the current — and

emittance, since its growth depends on compression — of the output beam

depends on the PWT injection phase. Shot-to-shot fluctuations of the rf phase

of more than a few degrees will cause significant fluctuations in the current

and emittance of the compressed beam.

In addition to providing focusing, the entrance and exit edge angles of

the chicane define the horizontal position of the beam axis (it is the position

where the length of the first magnet equals that of the second and third.) As

with all focusing elements, a misalignment of the beam results in steering.

46
This means that the magnets must be aligned to the beam axis horizontally

as well as vertically. It also means that the field of each magnet must be con-

trolled precisely and repeatably. To do this, each dipole has a small trim coil

to make minor adjustments to the magnetic field. In operating the Neptune

compressor, we found that 1% control over the field strength in each dipole

is required to eliminate steering the beam.

1.3.6 Beam Diagnostics

In view of the goal of producing high-brightness beams, and the fact that the

applications of such beams require accurate measurement of their parame-

ters, beam diagnostics are an important part of the photoinjector. In partic-

ular, the extreme self-fields of photo-beams and their shot-to-shot variability

are the important features considered in the implementation of diagnostics.

Of course, there are many measurements, not directly related to the beam

parameters, that must be made on a regular basis in operating a photoinjec-

tor, the rf phases being a prime example. Here, the term beam diagnostic

refers to a direct measurement of some aspect of the beam brightness. In

practice, diagnostics are categorized by the projection of brightness they

measure: charge, transverse emittance, bunch length, momentum and mo-

mentum spread, and as a sub-diagnostic (because it’s often a component of

another diagnostic, and is important in its own right) beam profile. The di-

agnostics used to measure emittance and bunch length are important to the

topics of this dissertation, and are each described in detail in Chapters 3

and 4, respectively. At Neptune a slit-based system is used to measure emit-

47
tance. The slit system addresses both of the major design considerations as

it is single-shot, and the effect of space-charge is mitigated in the measure-

ment by collimation of the beam into emittance dominated beamlets [52]. To

measure bunch length, a diagnostic based on coherent transition radiation

(CTR) is used [53, 54]. Unfortunately, this measurement scheme is not single

shot, and is affected by fluctuations in rf phase and beam charge. It does,

however, provide a direct measurement of the sub-picosecond bunch lengths

produced by the compressor.

The beam charge is measured using an integrating current transformer

(ICT) [55]. As its name suggests, the ICT is a ferrite-loaded transformer which

produces an output signal proportional to the charge of the beam passing

through it. The output signal is amplified and sent to a gated, charge col-

lecting analog-to-digital converter (ADC) where the signal can be read by a

computer. The range and precision of the measurement depends on the con-

figuration of the read-out electronics, which can easily be adjusted to mea-

sure charges from 30 pC to 10 nC. This is more than adequate for beams

produced at Neptune, and the typical configuration is set to read charges in

the neiborhood of 500 pC with a 10 pC resolution. The ICT measurement

is single-shot and non-destructive. This property allows the measurement

of charge and another (usually charge dependent) beam parameter, such as

emittance or spot size, concurrently.

One aspect of the bunch compressor not discussed so far, is that it may

be used as a spectrometer by removing the current in the first two magnets

and running the field in the last two in the same direction. In this config-

uration the beam energy and energy spread are measured by bending the

48
beam through the fixed radius of curvature defined by the spectrometer exit

port and intercepting it with a beam profile screen. This measurement is

single-shot because the intercepting screen measurement is single-shot, as


described below. The beam dispersion (ηx ≡ δx δp/p) at the measurement

screen and the screen’s resolution determine the momentum resolution of

the spectrometer. At full beam energy the minimum resolvable relative en-

ergy difference in this system is about 5 keV. The absolute energy calibration

is limited by the alignment of the spectrometer arm to about 5 %. The spec-

trometer arm serves a second function because the bend angle it requires is

the same as the design bend angle of the compressor. Thus, the current in

the dipoles can be properly set for compression by centering the beam on

the spectrometer profile monitor, regardless of beam energy.

The beam profile monitor, often called a screen, consists of a fluorescent

material that can be inserted into the beam path. The beam electrons excite

this material and the resulting light emitted is detected by some means. The

screens at Neptune come in two varieties of fluorescing material: Gd2 O2 S : Tb

deposited on a thin metal sheet [56], and 0.5 mm thick YAG:Ce crystals [57].

The YAG screens are preferred in practice because they produce more light

than the older style phosphor screens. The screens are mounted normal

to the beam axis on remote controllable actuators. The resulting light is

reflected through a window out of the vacuum pipe by a mirror mounted di-

rectly behind the screen at a 45◦ angle. The surface of the screen is imaged

by video camera located outside the vacuum window, and the video image of

the beam is sent to the control system for analysis. The resolution of beam

image from the screen is determined by the grain size of the fluorescing ma-

49
terial, the optics of the camera lens and the size of the light-sensing pixels

in the camera. Usually, the screen mount has fiducial markings so the po-

sition of the beam relative to the center of the beamline can be determined.

The camera’s field of view is therefore made large enough to include these

marks, and the resolution, here, is limited by the lens optics and pixel size.

For most of the view screens at Neptune the resolution is set to roughly 30µm

per pixel.

The beam profile monitor is an indispensable tool in the operation of a

photoinjector. This point is evident in the layout of the Neptune beamline

(Figure 1.8), where a view screen is located before or after (usually before and

after) every non-drift element. Operationally, they are the main diagnostic

used to propagate the beam down the beamline to the experiment. They are

used to examine the steering and focusing of the beam as it traverses the

beamline elements. If the light emitted by the fluorescer is linearly propor-

tional to the amount of charge it intercepts (this is true for both types of

screens described above), the screen is truly a profile monitor, and the image

contains information about the beam’s density distribution. This fact makes

the screen useful as part of a diagnostic for measuring beam parameters such

as rms size, emittance, and energy spread.

1.3.7 Control System

The heart of the Neptune control system is a Macintosh computer running

LabVIEW [58] measurement software. The computer has a GPIB [59] interface

card that allows it to collect data from a variety of sources such as oscillo-

50
scopes, CAMAC [60] modules, and other GPIB enabled hardware. The video

images from the view screens are sent to a frame-grabber card installed in

the computer. The advantages of the LabVIEW program are two-fold; it is de-

signed to make communication with external hardware a simple task, and it

has a built-in programming language which allows the user to automate mea-

surements (for example, taking a series of beam size measurements while

adjusting the focusing elements in the system) and manipulate the acquired

data. The video frame-grabber, the fast, GPIB-connected ADC used for the

ICT measurement, and (of course) the oscilloscopes in the control system are

all triggerable so that they collect data only when the beam is present.

While a complete description of the control system could fill a volume,

most of the technical details are of little relevance to the production of high-

brightness beams (they are required to run the machine, whether the beam

quality is good or not.) In this regard, the important feature of the hard-

ware described above is its ability to collect and analyze shot-to-shot data.

Specifically, since beam current (space-charge) is so important in photoin-

jectors, and since the ICT measurement is single-shot and non-destructive,

the control system is capable of collecting charge data along with another

parameter (such as beam size, emittance, rf phase, etc.) on a shot-to-shot ba-

sis. Furthermore, the data-analysis is usually preformed as the data is being

collected. These two features are a great aid to the operator of the machine

because they permit the otherwise complicated analysis of a measurement

to be done while the beam is still there. By iteratively doing this type of mea-

surement and adjusting the system settings, the operator can go through a

sort of “tuning process” to maximize the quality of the beam transported

51
through the system. Examples of this process discussed in latter chapters

are the measurement of emittance while adjusting to solenoid strength, and

measurement of pulse length versus PWT injection phase.

1.4 Chapter Summary

Modern electron beam applications require extremely high quality beams.

The beam’s quality may be defined by its six-dimensional phase space den-

sity, or brightness. The transverse component of brightness is emittance,

which is a measure of the area of the (x, px ) phase plane occupied by the

beam. The standard method of defining emittance in photoinjector physics

is through the root-mean-square beam size and divergence. This method has

the advantage that the evolution of the rms quantities does not depend on

the specific form of the distribution function. The emittance and brightness

of a beam are used as standard figures-of-merit to determine wether or not a

given beam is acceptable for a specific application, although often times an

application may place special demands on certain projections of the bright-

ness, such as current or energy spread.

Previous sources of electron beams, namely thermionic DC guns, were

found to be insufficient for producing the beam densities required by appli-

cations, and this was a strong motivation for the development of photoinjec-

tors. The photoinjector uses two key technologies to produce high brightness

beams: strong rf fields to accelerate the beam quickly to relativistic energy,

and photo-emission induced by pico-second lasers, to produce beams with

extremely high current densities and dimensions that are much smaller than

52
the rf wavelength. These technologies combine in the photoinjector to pro-

duce high current, low energy spread, low emittance beams, and have made

the photoinjector the preferred source for demanding applications.

The Neptune photoinjector is an example of the implementation of this

type of source. Each of the components of the system are designed and

operated in such a way that the beam quality and stability are maximized.

Where issues of stability do exist, for example laser positioning and PWT

injection phase, measurement and feedback systems are used to monitor and

control them. For that matter, diagnosis of the beam itself is very important

in minimizing emittance and propagating the beam. Because of this, the

Neptune photoinjector is very well diagnosed, with single shot measurements

preferred whenever possible. The beam current can be increased using a

magnetic compressor, whose compact design and sensitivity to rf phase also

require good beam stability and diagnosis.

53
 Chapter 2

Rectilinear Space-charge Dynamics

With an understanding of the goal, which is the production of the highest

brightness beam possible for applications such as SASE-FELs and advanced

accelerator experiments, the photoinjector is used because it is capable of

creating beams of intrinsically higher brightness than previous sources. A

commonly employed method of achieving this goal is as follows: Create, ac-

celerate, and transport the beam in a straight line in such a way that the

transverse emittance is minimized, then magnetically compress using a chi-

cane geometry to increase the beam current. The first part of this task is the

topic of this chapter; the second part is the topic of Chapter 4.

To minimize the emittance during acceleration and transport, the physics

of the problem must be understood. Because the beam is so dense, this

physics is dominated by the balance between external focusing and accelera-

tion forces and the repulsive space-charge force of the beam. Particle trajec-

tories in phase space are fundamentally different from those of thermalized

(e.g., bi-Gaussian) beams because of the central role space-charge plays in

photoinjector beam dynamics. This different behavior is a reflection of the

plasma nature of the beam, and as the analysis shows, the beam’s plasma

frequency, ωp , is a crucial parameter in its description. Perhaps the most

54
telling signature of the space-charge dominated behavior of photoinjector

beams is the emittance oscillations it undergoes (at the plasma frequency)

during acceleration and transport.

The analysis of the beam’s rectilinear space-charge dynamics can be bro-

ken down further into two different physical processes. The first concerns

the relative behavior of different longitudinal slices of the beam. The theory

describing this process is termed linear space-charge emittance compensa-

tion, and was first introduced by Carlsten [61], and latter developed further by

Serafini and Rosenzweig [62]. Emittance compensation theory is paramount

in the design and operation of photoinjectors, and has been verified by sim-

ulation and experiment [63]. Emittance compensation is a linear theory be-

cause it deals with the effects of the linear component of the space-charge

and external forces on individual beam slices. The crux of linear emittance

compensation is that the space-charge force on a slice is determined by the

density profile of the beam and the position of that slice within the beam.

As a result, different slices are exposed to different forces, and therefore

expand at different rates and make different angles in phase space. As the

slices separate in phase space angle, the rms emittance of the full beam in-

creases. Emittance compensation, then, is the process of undoing the effect

of space-charge on the phase space angles of the slices by careful use of

external focusing to line the slices back up, reducing the emittance.

Note that even though the emittance changes in this process, Liouville’s

theorem is not violated. Actually, Liouville’s theorem does not strictly ap-

ply to the (x, px ) phase plane since the transverse space-charge forces on

a slice are coupled (through the longitudinal charge distribution) to that

55
slice’s position in z. It does apply however, to the beam’s full six dimen-

sional phase space. In fact, if one considers the three dimensional space

defined by (x, px , z), the beam in this space resembles a ribbon stretched out

along the z-axis. The thickness of this ribbon is the thermal emittance of the

longitudinal slice, and does not change in the linear emittance compensation

theory. If this ribbon is twisted, the area occupied by its projection onto the

(x, px ) plane (its projected emittance) changes, even though the volume of

the ribbon remains the same.

The other process is, in a way, the opposite of the first. It is the effect

of non-linear space-charge fields within a single longitudinal slice. Nonlinear

fields occur when the transverse charge distribution is not constant. When

this happens, the phase space correlation of a slice becomes nonlinear, and

emittance growth results. The slice emittance can oscillate in this process as

well, with the oscillation again parameterized by the plasma frequency. The

physics of emittance growth due to nonlinear, transverse space-charge has

been studied in the context of intense proton and heavy ion beams [64]. For

photoinjector beams, the physical process is the same as with ion beams, but

the application of the result — that is, the best way to manipulate the beam,

such that the emittance growth is minimized — is different. This difference is

due to the relative lifetimes of the different beams, measured by the number

of plasma oscillations they undergo. With photo-beams this number is usu-

ally limited to only a few, and the long time scale effects of this process do

not have the opportunity to develop. As the analysis will show, the nonlinear

space-charge forces cause phase space wave-breaking in the beam. The in-

crease in emittance cannot be fully removed after the point of wave-breaking.

56
The emittance will oscillate after wave-breaking, however, and this oscillation

can be exploited (just as it is in linear emittance compensation) to minimize

the emittance at a given point in the beamline.

This chapter examines both linear emittance compensation theory and

the process of emittance growth and wave-breaking due to nonlinear space-

charge. In both cases the objective is the minimization of the projected rms

emittance of the beam.

2.1 Linear Space-charge Emittance Compensation

2.1.1 Space-Charge Dominated Behavior of Beam Slices

Much of the physics of this chapter can be understood by investigating the

behavior of a single beam slice when subject to space-charge forces but no

external forces. This is true because the slices are uncorrelated in the sense

that the dynamics of one slice do not directly effect the dynamics of the oth-

ers. This is the method of analysis originally used by Carlsten in introducing

the concept of emittance compensation. The model calculation, given below,

is meant to approximate the behavior of the beam emerging from the photo-

cathode. In this model, the beam is created with no phase space correlation,

and a small thermal velocity spread, which is ignored, as it has negligible im-

pact on the beam dynamics. The beam expands from the initial waist due to

the dominant space-charge forces, and different longitudinal slices expand

at different rates because the magnitude of the forces varies between slices.

After its initial expansion, the beam is focused back down to a small size

57
using a thin lens. The converging slices after the thin lens are also affected

by the differing space-charge forces, and this is the basic principle of emit-

tance compensation. The model is meant to be illustrative; it is by no means

a complete description of the problem (there is no acceleration), but it does

reveal the fundamental space-charge dynamics that dominated the slice mo-

tion after the photo-cathode.

To calculate the space-charge force, a cylindrically symmetric, long beam

(two dimensional) limit is assumed. This limit is reached when the beam

is highly relativistic (γ  1), where the rest frame bunch aspect ratio (A =

σr /γσz ) is small. Note that in the rest frame even short pulses appear elon-

gated because of length contraction by γ in the lab frame. The radial electric

field in this case is determined to an excellent approximation by the charge

density of the beam slice

nb (ζ, r , z) = λ (ζ) f (r , z) , (2.1)

where λ (ζ) is the number of electrons per unit length, dependent on the

position of the slice within the bunch, ζ = z − ct, (the current of the slice

is therefore given by I (ζ) = evz λ (ζ)  ecλ (ζ)) and f (r , z) is the radial

density, which will vary with z as the beam is defocused. For the purpose of

doing the linear analysis, a constant radial density with a hard edge at r = a,

f (r , z) = 1 π a2 (z), is assumed. The electric field inside the slice is found,

therefore, using Gauss’s law

eλ (ζ) r
Er (ζ, r , z) = · 2 , (2.2)
2π 0 a (z)

where 0 is the permittivity of free space. The field is linear in radius, but

depends on the radial extent of the slice, a, which is a function of z.

58
 To go farther, the concept of laminar flow, which will be quite useful in

the analysis of nonlinear space-charge fields, can be used to find a (z) in

terms of r (z). Laminar flow imposes the condition that particle trajectories

do not cross. This means that the number of particles within a set fraction

of the beam radius, r /a, is constant, or

r  r0
2π r̃ nb (r̃ ) dr̃ = 2π r̃0 nb (r̃0 ) dr̃0 , (2.3)
0 0

where r0 = r (z = 0). The validity of the laminarity condition will be con-

firmed later using the single particle equations of motion (if particles never

cross, the condition dr /dr0 ≠ 0 must alway be true). If Equation (2.3) holds,

it can be evaluated easily in this case, giving the relationship between r and

a
a20
a2 (z) = r 2 (z) . (2.4)
r02
The field is therefore
 2
eλ (ζ) r0 1
Er (ζ, r , z) = . (2.5)
2π 0 a0 r (z)

The beam current creates an azimuthal magnetic field, which is calculated

using Ampere’s law and the above laminarity condition

 2
eβz cµ0 λ (ζ) r0 1 βz
Bθ (ζ, r , z) = = Er . (2.6)
2π a0 r (z) c

The radial motion of a particle is found using these two field components in

the Lorentz force law

d
   eE
r
pr = e (Er − βz cBθ ) = eEr 1 − β2z  2 . (2.7)
dt γ

As advertised in Chapter 1, the space-charge force is proportional to 1/γ 2 ,

and it is clear now that this comes from the difference in magnitude between

59
the defocusing electric and focusing magnetic fields of the beam. To use z

as the independent variable instead of t, note that

d
 d d2 r d2 r 2
2d r
pr = (γmvr ) = γm 2 , and = v z  β2 c 2 r

, (2.8)
dt dt dt dt 2 dz2

where γ is moved outside the time derivative because there is no external

acceleration. Using Equations (2.5), (2.7), and (2.8), the single particle motion

is given by
 2


e2 λ (ζ) r0 1
r (ζ, z) = . (2.9)
2π 0 β2 γ 3 mc 2 a0 r (z)

The collection of physical constants is simply the classical electron radius

e2
re = , (2.10)
4π 0 mc 2

and the transverse plasma wavenumber (spatial plasma frequency), kp =

ωp /c, defined by
4π re nb (ζ, z)
k2p = , (2.11)
β2 γ 3

can be used to simplify Equation (2.9), using k2p (ζ, z = 0) = k2p0 (ζ), to

1 2 1
r

(ζ, z) = kp0 (ζ) r02 . (2.12)
2 r (z)

Unfortunately, this equation has no exact solution, but may be reduced

to an integral equation, as shown by Reiser [3]. To do this, it is useful to

introduce a change in variables.

r dR r

R= , Z = kp0 z, R
= = . (2.13)
r0 dZ kp0 r0

Using these new variables, Equation (2.12) takes the form

1
R

= . (2.14)
2R

60
Following Reiser’s analysis of this equation leads to

 R0
±ln R
2

−R0

2
2
Z = 2e eR dR
, (2.15)
R0

where the minus sign in the upper limit applies when the beam is converging

(R
< 0). If the slice is initially at a waist, this equation has an approximate

solution

Z  2 R − 1, (2.16)

which is quite accurate until the beam size doubles. Solving for R and substi-

tution of the original variables gives the equations of motion for the particles

in the slice,

1 2
r (ζ, z) = r0 + k (ζ) r0 z2 (2.17)
4 p0
1
r
(ζ, z) = k2p0 (ζ) r0 z. (2.18)
2

Notice that the assumption that the beam size remains roughly constant

is implicit in this approximate solution (replacing r (z) with r0 in Equa-

tion (2.12) gives the same solution). Note also, that the solution is consis-

tent with the laminarity condition because dr /dr0 = 1 + 14 k2p0 (ζ) z2 > 0.

These equations show that the expansion rate of the slices is proportional to

k2p0 (ζ), and therefore to λ (ζ) and the current of the slice, I (ζ).

To visualize the beam expansion, consider the specific case of a Gaussian

longitudinal charge density

ζ 2
N − 2
λ (ζ) = √ e 2σz . (2.19)
2π σz

The maximum plasma wavenumber occurs in the middle of the bunch, kp0,max =

kp0 (ζ = 0), and Equations (2.17) and (2.18) can be rewritten with the ζ de-

61
 2.5 1.2
ζ=0
2 1

σz
0.8

r
kp0,max a0
1.5
r r0

2σz 0.6


1
0.4


0.5
0.2 (b)
(a)

0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5

kp0,max z r a0

Figure 2.1: (a) The expansion of different slices in a beam with a Gaussian
longitudinal profile. (b) The trace space positions of the slices after expan-
sion.

pendence explicit,

ζ 2
1 2 − 2
r (ζ, z) = r0 + kp0,max e 2σz r0 z2 (2.20)
4
ζ2

1 2 − 2
r (ζ, z) = kp0,max e 2σz
r0 z. (2.21)
2

These equations are used in Figure 2.1 to plot the expansion of slices at

ζ = 0, σz , and 2σz , and the trace space of the beam after the expansion.

The full trace space of the beam shows the “bow-tie” shape, characteristic of

photoinjector beams. The projected emittance — that is, the rms emittance

of the full beam — has increased even through the emittance of any given

slice has not (the slice emittance is zero in this model).

Now consider the slice dynamics if a thin lens focusing element, with focal

length f , is applied at position zl . Taking z to now represent the drift length

62
after the lens, the particle position is given by

rl 1 2
r (ζ, z) = rl + rl
− z+ kp0 (ζ) r0 z2 , (2.22)
f 4

where rl and rl
are the position and angle of the particle given by Equations

(2.17) and (2.18) at the position of the lens, zl . With some algebra, this equa-

tion simplifies to
 
1 2 z 1 2
r (ζ, z) = r0 + kp0 (ζ) r0 (z + zl )2 − r0 + kp0 (ζ) r0 zl2 . (2.23)
4 f 4

Similarly, the r
equation becomes

rl 1
r
(ζ, z) = rl
− + k2p0 (ζ) r0 z
f 2
 
1 2 r0 1 2 2
= k (ζ) r0 (z + zl ) − 1 + kp0 (ζ) zl . (2.24)
2 p0 f 4

If the thin lens transformation is applied to the trace space of Figure 2.1,

the trace space in Figure 2.2 results (in this case f was chosen to make the

highest current slice flip its angle, r
→ −r
).

Flow is again laminar under Equation (2.23) because all of its terms are

proportional to r0 . In this case the laminarity condition, dr /dr0 = r /r0 > 0,

states that the above expressions are invalid if r becomes negative (the point

that r cannot be negative is physically obvious, but it illustrates the mathe-

matical machinery which is needed in the analysis of nonlinear space-charge).

Physically, the slice trajectory is initially converging, and will evolve into one

that is diverging. This can happen in two different ways: the first in which

r is always positive and governed by Equation (2.23). Here the minimum in

r occurs when r
is zero for every particle. In phase space the slice rotates

counter-clockwise and crosses the r axis. The other case is the one in which

63
 0

-0.5

-1

r
kp0,max a0 ζ=0
-1.5


σz

-2
2σz

-2.5
0 0.5 1 1.5 2 2.5

r a0

Figure 2.2: The trace space of the beam in Figure 2.1 after a thin lens trans-
formation.

r goes to zero for every particle and Equation (2.23) becomes invalid. This

time the slice rotates clockwise in phase space and aligns with the r
axis,

after which the sign of each particles angle flips and the beam becomes di-

vergent. These two different situations are illustrated in Figure 2.3. When

the space-charge forces are sufficiently strong, the first case will occur, and

the particles will cross the r axis. In this situation the beam is said to have

formed a space-charge dominated waist. If the space-charge fields are not

strong enough to force a waist, the minimum beam size is referred to as a

ballistic waist, or a cross-over, as termed by Carlsten. To determine how

space-charge controls which type of waist occurs, the laminarity condition

64
 x

waist
x

cross-over

Figure 2.3: Illustration of the two possible trajectories of a beam slice. A

waist occurs when the slice crosses the x axis, a cross-over when it crosses
the x
axis.

can be examined to see under what conditions it is violated.

    
dr 1 2 2 1 2 1 1 2 2 1
= 0 = 1 + kp0 zl + z kp0 zl − 1 + kp0 zl + k2p0 z2 (2.25)
dr0 4 4 f 4 4

Here the equation is grouped in powers of z and the explicit ζ dependence of

k2p0 is left out for convenience. The equation has the quadratic form ax 2 +

bx + c = 0, and therefore has real roots — and Equation (2.23) becomes

invalid — if the quantity b2 − 4ac is positive. Using this, we can find the the

minimum focal length that avoids a ballistic waist by solving b2 − 4ac = 0,

65
 3.5

3
cross-over

2.5

2
zl fmin

waist


1.5

0.5

0
0 1 2 3 4 5

kp0 zl

Figure 2.4: Plot of the minimum cross-over focusing strength as a function of

the slice plasma wavenumber. Slice that lie above the curve will go through
a cross-over.

or
  2  
1 2 1 1 2 2 2 1 2 2
k zl − 1 + kp0 zl − kp0 1 + kp0 zl = 0. (2.26)
4 p0 f 4 4

From this the minimum focal length is

 
1 1 2 3 1 1 2 2 3/2
fmin = − zl − kp0 zl + 1 + kp0 zl . (2.27)
2 8 kp0 4

The maximum focusing strength of the lens, 1/fmin , increases with the plasma

wavenumber of the slice, as shown in Figure 2.4. If the minimum focal length

for a given slice is used, the position of the beam waist is found by substi-

66
tuting Equation (2.27) into Equation (2.25).
 

2 1 2 2  4
z= 1 + kp0 zl = zl2 + 2 (2.28)
kp0 4 kp0

All of this analysis is leading to the identification of a problem. That is, in

order to sufficiently focus the beam slices with the most current, those slices

with the least current are given a steep angle in phase space, and are likely

to go through a cross-over waist, while the high current slices go through a

gentle, space-charge dominated waist. The way to avoid this bifurcation of

the beam’s phase space is to minimize the range of current values between

slices. Specific expressions for the limits of the current range can be found

by examining the behavior of the slices that go through a space-charge dom-

inated waist after the thin lens.

Assuming, for the moment, that for some set of slices the laminarity con-

dition always hold and that Equations (2.23) and (2.24) are valid, the emit-

tance of that set of slices will begin to decrease with z. To see this, we examine

how r and r
vary with k2p0 .

 
1 2 zzl2 r0
∆r = r0 (z + zl ) − ∆k2p0 ∝ z2 , (2.29)
4 4f
 

1 zl2 r0
∆r = r0 (z + zl ) − ∆k2p0 ∝ z. (2.30)
2 4f

The difference in phase space angle, and therefore the emittance, decreases

as z increases, since
 
r
∆r
· r − ∆r · r
∆k2p0
∆ = ∝ . (2.31)
r r2 k2p0 z

It can be seen from this that the emittance only decreases because the plasma

frequency varies between slices — the same reason it increased before the

67
thin lens. In fact, if ∆r
· r − ∆r · r
= 0, the phase space angles of the slices

will be equal and independent of kp0 , and the emittance will vanish. This

compensation point, where the oscillation in emittance is completed, occurs

when
1 z + zl
=2 . (2.32)
f z2

This focal length must be greater than the cross-over focal length, and

this condition, combining Equations (2.27) and (2.32), gives

 
z2 1 1 2 3 1 1 2 2 3/2
≥ − zl − kp0 zl + 1 + kp0 zl , (2.33)
2 (z + zl ) 2 8 kp0 4

or
2
k2p0 ≥ . (2.34)
z2 − zl2
From this, compensation can not occur unless z > zl , which can be seen from

Equation (2.28), since the position of the cross-over is always greater than zl .

The restriction on k2p0 can be quantified further by recalling that this analysis

uses the approximate equation of motion given by Equation (2.16), which is

accurate as long as kp0,max zl  2. For the same reason, z should be made as

small as possible while staying above the limit given by Equation (2.28), and

Carlsten suggests that this limit should not be exceeded by more than about

10%. Using kp0,max zl = 2 in Equation (2.28) and a 10% buffer gives z ≈ 1.56zl ,

and using this in Equation (2.34) gives a condition for the allowable variation

in k2p0 ,

k2p0 (ζ)  0.35k2p0,max . (2.35)

Thus, under these conditions, slices with less than about one third the max-

imum density have too little current to oscillate with the core of the bunch,

and will experience cross-over waists.

68
 2 8

6
compensation
1.5 4 point
2

r
r
r r0


1 0


-2

0.5 -4

-6

0 -8
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
 
z zl z zl

Figure 2.5: (a) Radius of four

 slices before and after the thin lens. (b) The


phase space orientation, r r , of the slices is plotted versus z. The focusing
was chosen to produce the compensation point at z ≈ 1.56zl after the lens.

The process is plotted in Figure 2.5 for 4 slices with z, zl , and f given

by the above analysis. As mentioned above, this model is an approximation

of the behavior of the beam from the cathode through the rf gun and fo-

cused by the (emittance compensating) solenoid. The model illustrates the

important features of the process by examining the space-charge forces on

different longitudinal slices. Here, the emittance undergoes an oscillation —

a feature that will turn out to be a general property of space-charge domi-

nated beam behavior. This oscillation is used in photoinjectors to recover

the high brightness of the beam that exists at the cathode. In this case, the

oscillation is adiabatically terminated by positioning an accelerator section

(like the PWT at Neptune) near or at the compensation point and accelerat-

ing the beam with low emittance — that is, rapidly increasing γ to lower the

space-charge forces where slices align in phase space. The model also shows

69
that the emittance oscillation is due to variations in the space-charge forces

and estimates the size of these variations for which the behavior of different

slices will be dramatically different.

2.1.2 Envelope Analysis

In order to perform a more general analysis of the behavior of beam slices,

the envelope equation formalism is needed. As will be shown here, the use

of envelope equations allows the inclusion of a simple, linear model of all

the relevant physical effects, such as space-charge forces, external focusing

and acceleration forces, and emittance, in a straightforward manner. The

problems encountered in the following sections of this chapter are based on

finding and solving relevant envelope equations, and, therefore, the general

rms envelope equation is derived here.

To derive an rms envelope equation, one differentiates the rms beam size,
 

d xx

σx ≡ x  =
2 , and (2.36)
dz σx
 
1 dxx
 xx
2
σx

= 2 σx − (2.37)
σx dz σx

If acceleration is to be considered, one must be careful to remember that the

angle x
≡ px /ptot in evaluating xx

, which implies that (x
)
= px
/ptot −

2
px ptot /ptot . Here the first term is due to transverse forces (either internal

or external) applied to the particles, while the second term arises because x

must decrease as the beam accelerates and ptot increases. With this in mind,

Equation (2.37) is written as





2
1 γ xx 
σx

= 2 σx x
2  + xx

 − xx
 − , (2.38)
σx γ σx

70
where the ultra-relativistic limit is assumed so that (βγ)
 γ
. With some

straightforward algebra, this reduces to

γ

ε2 xx


σx

+ σx = x3 + , (2.39)
γ σx σx

where εx is the un-normalized rms emittance given by the standard defini-

tion, εx2 = x 2 x
2  − xx
2 . The xx

 term represents the effect of all

of the transverse forces in the problem, and for the analysis that follows, it

will be sufficient to specify this term for three types of forces: space-charge,

linear external focusing forces from magnetic optics, and the (averaged over

one period in z) alternating-gradient focusing of the rf fields in an accelerator.

In the case of linear external focusing, the force term is easy to evaluate,

because the equation x

= −k2β (z) x can always be written, where k2β (z) is

used to indicate that the force can be piece-wise continuous in z. In the case

of quadrupole focusing, k2β = eB
/ptot over the effective length of the lens.

With this form for x

, the force term in Equation (2.39) is

xx

 k2β x 2 
=− = −k2β σx (2.40)
σx σx

If the beam and the forces in the problem are both axially symmetric, the

equation for the radial rms envelope can be written simply by replacing x

with r in Equation (2.39), and defining the radial emittance as εr2 = r 2 r
2 −

r r
2 , where r
= pr /ptot . This symmetry applies when the beam is focused

by a solenoid, and kβ = eBz /2βγmc 2 .

The radial equation is the appropriate choice when considering the fo-

cusing due to rf accelerating structures because here again the forces are

71
axisymmetric. In this case it has been shown [30] that the force is given by

er dEz
Fr  − , (2.41)
2 dz

and the average force is

η (eE0 )2
Fr  = − · r, (2.42)
8 γmc 2

where E0 is the peak accelerating field in the structure, and η is a unitless

factor that depends on the spatial harmonics of the rf fields. Using Equa-

tion (2.8) and dropping factors of β, the equation for r

is

2
2


η eE0 η γ

r =− r =− r, (2.43)
8 γmc 2 8 γ

where the normalized energy gain is given by γ
= eE0 /mc 2 . The r r

 term

in the radial envelope equation due to rf focusing is therefore

2
r r

 η γ

=− σr (2.44)
σr 8 γ

Notice that solenoid focusing has the same γ dependence as rf focusing,

and that in the case where both are present, the solenoidal focusing may

be included in this term by making the substitution η → η + 2b2 , where

b = Bz /E0 .

At this point the radial envelope equation including focusing and accel-

eration may be written as

2
γ
η γ
εr2 r r

sc
σr

+ σr
+ σr = + . (2.45)
γ 8 γ σr3 σr

The term left to be evaluated, r r

sc , is due to space-charge forces. If the

radial slice density is a constant, then we have found that the space-charge

72
force is linear in r , and Equation (2.12) gives r r

as

1 2 2re λ (ζ)
r r

= k r = 2 3 2 r 2. (2.46)
2 p0 β γ a

Here, as with the other terms in the envelope equation, the long beam, ultra-

relativistic limit is assumed, so that the force depends only on the slice den-

sity per unit length, λ (ζ), and γ. Likewise, the factor of β2 can be dropped,

and the average over the the beam is found by

a
2π r (r r

) nb (ζ, z) dr
r r

 = 0
a , (2.47)
0 2π r nb (ζ, z) dr

from which the space-charge term is found to be

r r

 re λ (ζ)
= or, (2.48)
σr γ 3 σr
I (ζ)
= , (2.49)
γ 3 I 0 σr

in terms of the current of the slice.

It turns out that this form of the space-charge term is valid for any density

profile, nb (ζ, r , z). This is true, as shown by Sacherer [7], because the rms



beam size depends only on the linear part of the space-charge force, rlin =

Flinr . Here, Flin is defined in the least-squares sense, by minimizing the

quantity

2
∆= 2π r [Flinr − r

] nb (ζ, r , z) dr . (2.50)

This difference is minimized, d∆/dFlin = 0, when

Flinr 2  = r r

, (2.51)

which is the same as saying that r r

lin = r r

, so Equation (2.49) holds

for any radial charge profile.

73
 The radial envelope equation, including space-charge, acceleration, exter-

nal forces, and emittance, is therefore,

2
γ
η γ
εr2 I (ζ)
σr

+ σr
+ σr = 3 + 3 . (2.52)
γ 8 γ σr γ I 0 σr

If the beam does not have axial symmetry, envelope equations in x and y

must be used, but this is not a great difficulty since it has already been shown

that going from r to x or y is little more than a matter of replacing the

subscripts in most of the terms in the above equation. Although it was not

explicitly shown here, this substitution is also valid for the term due to rf

focusing [29]. The space-charge term, however, requires more modification.

Again, in the continuous beam limit, it can be shown [7] that if the distribution



has elliptical symmetry — so that nb ζ, x, y, z = nb ζ, x 2 /a2 + y 2 /b2 , z ,

where a and b are the semi-major axes of the ellipse — then the space-charge

term becomes
I (ζ)

,
γ3I 0 σx + σy

which is again good for any transverse charge distribution. Therefore, the x

envelope equation is

2
γ
η γ
εx2 I (ζ)
σx

+ σx
+ σx = 3 + 3

, (2.53)
γ 8 γ σx γ I 0 σx + σy

with an analogous equation in y. The remainder of this chapter is concerned

with finding solutions to Equation (2.52), while in Chapter 3 the x and y equa-

tions will be important in determining the effect of space-charge on emittance

measurements.

74
2.1.3 Brillouin Flow

The generalized theory of emittance compensation, formulated by Serafini

and Rosenzweig [62] (SR), examines the solution of Equation (2.52) for differ-

ent beam slices. The analysis takes the cold-fluid limit, in which the emittance

term in the envelope equation is much smaller, throughout the course of the

beam’s motion, than the space-charge term, and is therefore dropped. This

limit, in which the ratio of the space-charge and emittance terms is much

greater than one, defines the concept of a space-charge dominated beam. In

this limit, and in the case where there is no acceleration and the beam is

focused by a solenoid, the radial rms envelope equation becomes

re λ (ζ)
σr

(ζ, z) + k2β σr (ζ, z) = 3
, (2.54)
γ σr (ζ, z)

where the ζ dependence is included to stress the fact that the solution will

be different for each beam slice. As with Equation (2.12), there is no exact an-

alytical solution to this equation, but there is an equilibrium solution, where

σr

= 0 
1 re λ (ζ)
σeq (ζ) = . (2.55)
kβ γ3

The steady state envelope given by this equilibrium corresponds to a rigid

rotor equilibrium known as Brillouin flow [65], in which the beam’s canonical

angular momentum is zero. The typical way of dealing with Equation (2.54)

is to expand it to first order about σeq in the parameter δσr = σr −σeq  σeq .

The linearized envelope equation becomes

re λ (ζ)
δσr

(ζ, z) + k2β δσr (ζ, z) = − δσr (ζ, z) , (2.56)
γ 3 σeq
2 (ζ)

75
or

δσr

(ζ, z) + 2k2β δσr (ζ, z) = 0. (2.57)

If we again assume that all slices are initially launched with the same rms

size σr (ζ, 0) = σr 0 , and with no rms radial motion σr
(ζ, 0) = 0, then the

solution for small amplitude motion about the equilibrium associated with

each slice is

   
σr (ζ, z) = σr 0 + σr 0 − σeq (ζ) cos 2kβ z , (2.58)

with derivative
    
σr
(ζ, z) = − 2kβ σr 0 − σeq (ζ) sin 2kβ z . (2.59)


In this case the σr , σr
trace space trajectory of a slice envelope is simply


an ellipse whose origin is offset to σeq (ζ) , 0 . The mismatched envelopes
√
rotate about this offset position with wave number k = 2kβ , but according

to Equation (2.55) this is


 
 2re λ (ζ)
2kβ =  3 2 , (2.60)
γ σeq

which is just the plasma wave number, kp = 4π re nb /γ 3 , of the equivalent
1 2
2
uniform density matched beam, since σeq = a .
2 0
In the small amplitude

limit, the oscillation frequency is independent of λ (ζ). Thus every trajectory

of this form aligns in trace space twice per plasma period, points at which the

projected rms emittance of the ensemble of beam slices in this trace space,
   2
εr = σr2 σr
2 − σr σr
, (2.61)

vanishes. This definition of emittance is identical to that of the standard ra-

dial rms emittance if each slice is a line in (r , r
) trace space, which connects

76
 σr

λ1 λ2 λ3
σr

σeq1 λ1 < λ2 < λ3

σeq2 σeq3


Figure 2.6: Trace space trajectories for σr , σr
in a system launched with
size below the equilibrium for three representative slices,
 with √
line charges

λ1 < λ2 < λ3 . Oscillations proceed at the same frequency kp = 2kβ about
different equilibrium values of σr .


the origin to the edge of the slice distribution through the value σr , σr
.

This case, which is physically realized when the beam’s density distribution
√
is uniform out to a hard edge radius of a = 2σr , was the subject of the

envelope dynamics analyzed by SR.

In the case most relevant to the emittance compensation process, the

beam is launched with a size smaller than equilibrium for all portions of the

beam, and the trace space trajectories for various slices are nested ellipses.

This is shown in Figure 2.6, which displays three elliptical trajectories cor-

responding to three different slices with λ1 < λ2 < λ3 . These ellipses are

traversed in the linear analysis with the same frequency. Thus the area in

77
 σr

π
kp z = 2

λ1 λ2 λ3
σr

λ1 < λ2 < λ3
3π
kp z = 2

kp z = 0, 2π

Figure 2.7: Projected trace space areas described by the three slices of Fig-
ure 2.6, at kp z = 0, π2 , 3π
2 , 2π . Note the area (emittance) is maximized at
π 3π
kp z = 2
, 2 and vanishes at kp z = 0, 2π and also at kp z = π (not shown).

trace space that the points on the three ellipses describe when connected to

the trace space origin (at a given time), which is proportional to the emittance

defined by Equation (2.61), oscillates with twice the mismatch oscillation fre-

quency. This phenomenon is illustrated in Figure 2.7, which displays the

trace space area described by the three slices at kp z = 0, π2 , 3π

2 , 2π . It can

be seen that the trajectories fan out to produce a large projected emittance
π 3π
at kp z = 2, 2 , while to lowest order the emittance vanishes at kp z = 0, 2π

and also at kp z = π (not shown). These emittance oscillations repeat twice

every plasma oscillation, but eventually decohere due to small, amplitude

dependent differences in the plasma frequency in each slice arising from the

nonlinearity of the oscillation [66]. The proper execution of such an emit-

78
tance oscillation due to differential slice motion is termed emittance com-

pensation in the context of high current, space-charge dominated beams in

rf photoinjectors. This simple picture is complicated somewhat by acceler-

ation, as discussed below, but essentially illustrates the relevant physics of

compensation process.

The picture of the slice dynamics displayed in the trace space diagrams of

Figures 2.6 and 2.7 assumes — as is true of motion originating at a cathode

in an rf photoinjector — that the beam expands from its initial size, exceeds

an equilibrium value, and finally returns to its initial state. As this is not

the most general case, a more complicated, but equally relevant, picture is

displayed in Figure 2.8, where only two of the slices are launched with sizes

below equilibrium, but the third has low enough line charge density that, at

the same initial size of the other two slices, it is above equilibrium. This

picture displays what happens if a beam is launched with size matched in an

rms, integrated beam sense, so that all slices are the same size, but due to

variations in current, some slice sizes are initially above, and others below,

equilibrium. It can be seen that, while the slice dynamics and associated

emittance evolution are in some ways different (the maximum emittance is

larger in this case), the overall periodicity of the emittance oscillation is the

same. The most important way in which the two situations differ is that

in Figure 2.6 the rms beam angle σr
is the same sign for all slices, while in

Figure 2.8 the angle of the low current slice is of opposite sign from the other

two. This important point will be revisited below.

79
 σr

π
kp z = 2

λ1 λ2
σr

λ3
λ3 < λ1 < λ2

σeq1

σeq3 σeq2

Figure 2.8: Projected trace space areas described by the three slice envelopes
with line charges λ3 < λ1 < λ2 with the line charge of slice 3 so low that
π
σr 0 < σeq , shown at kp z = 2
. The emittance evolution behavior is qual-
itatively the same as in Figures 2.6 and 2.7, but with larger amplitude of
oscillation.

2.1.4 The Invariant Envelope

The extension of the above type of motion about an equilibrium to a system

with longitudinal acceleration has been considered by SR, who have analyzed

the motion of such a system with Equation (2.52) (but again, excluding the

emittance term). This equation is again nonlinear, but also has a useful par-

ticular solution — which is no longer an equilibrium, however — with which

one can begin an analysis, termed the invariant envelope [62],


2 re λ (ζ)
σinv (ζ, z) =
. (2.62)
γ (1 + η/2) γ (z)

80
It can be seen that the existence of this particular solution is not dependent

on external focusing, as even with η = 0 (pure traveling wave, no solenoid)

the state corresponding to this solution exists due to the effects of adiabatic

damping.

From the derivative of σinv ,



1 re λ (ζ)
σinv =− , (2.63)
γ (z) (1 + η/2) γ (z)

we see that the invariant envelope has the unique property that the trace

space angle, σr
/σr = −γ
/2γ, is independent of λ (ζ). Thus if one places all

slices on their invariant envelope, they will be aligned in trace space angle and

the emittance vanishes for all future times. It is not possible in practice to

do this, and so one must consider what happens when all slices in the beam

ensemble are placed close to their invariant envelopes. First, we examine the

motion of a slice perturbed slightly off of its invariant envelope, by using a

linear expansion of Equation (2.52) (without the emittance term) about this

particular solution,


2
γ
1+η γ

δσr

+ δσr
+ δσr = 0, (2.64)
γ 4 γ

where δσr = σr − σinv . This equation has a general form of solution, for the

type of initial conditions being considered here, of



1+η γ0
δσr = [σr 0 − σinv ] cos ln , (2.65)
2 γ (z)

so that we can write



1+η γ0
σr (z) = σinv + [σr 0 − σinv ] cos ln , (2.66)
2 γ (z)

81
and
 

1 + η γ
1+η γ0
σr
(z) = [σr 0 − σinv ] sin ln , (2.67)
2 γ (z) 2 γ (z)

with γ0 = γ (0). Thus the mismatch envelope dynamics are not conceptually

much different than in the coasting beam case, with oscillations about the

particular solution (no longer an equilibrium, but a secularly diminishing

envelope) proceeding at approximately the plasma frequency. Note that the

plasma frequency is no longer a constant in this case, but monotonically


−1 −3/2
diminishes with acceleration, approximately as kp ∝ nb /γ 3 ∝ σinv γ ∝

γ −1/2 . Mismatch envelope dynamics corresponding to Equations (2.66) and

(2.67) are illustrated by the normalized trace space (phase space) picture

given in Figure 2.9, which shows the dynamics of three slices corresponding

to the hierarchy of currents introduced in Figure 2.8.

While the picture in Figure 2.9 gives a similar schematic view of emit-

tance oscillations as Figure 2.8, it has two notable differences with the non-

accelerating case. The first is simply that the emittance one needs to be con-

cerned with when the beam accelerates is the normalized emittance, εr ,n 


γ σr2 σr
2  − σr σr
2 , which is a measure of the transverse phase space

area, and is thus conserved under linear transport and acceleration. The

usual “adiabatic damping” of the trace space area, in which the trace space

angles diminish through acceleration as (βγ)−1 , is emphasized in Figure 2.9

by rescaling of the vertical axis with γ (recall that we have set β = 1 in this

analysis). This rescaling removes the apparent damping of the motion and

makes the diagram approximately a correct phase space plot, in the limit

β = 1. The second difference is that all mismatch oscillations have end

82
 γσr

1
Area after 4 plasma oscillation.

σr

λ1 λ2

λ3

γσr
γ

σr =−2

Figure 2.9: Normalized, projected trace space areas described by three slices
with line charge λ3 < λ1 < λ2 as the envelopes oscillate about the individual
invariant envelopes, with the line charge of slice 3 so low that σr 0 > σinv .

points attached to a line γσr
/σr = −γ
/2 instead of σr
= 0. As the invariant

envelope associated with the slices becomes smaller with increasing energy

as γ −1/2 (the ensemble of ellipses shown slides up the line γσr
/σr = −γ
/2

towards the origin), the area associated with the emittance not only oscillates,

but secularly damps as γ −1/2 .

Note that the offset phase space area described by the mismatch oscilla-

tions (the ellipses in Figure 2.9) is actually conserved, as can be seen through

inspection of Equations (2.66) and (2.67). This means that for an ensem-

ble of slices placed all at the same initial phase space condition, but with

different λ (ζ), the set of points which makes up the section of the phase

83
space boundary not attached to the origin form a line with varying length

but no area. This ensemble line stretches and rotates about the invariant

envelope of the matched slice. If the invariant envelope slice is actually

present in the beam, the ensemble line passes through the invariant enve-

lope line, γσr
/σr = −γ
/2, and rotates about the intersection point of these

two lines. This intersection is therefore a fixed point in phase space. Thus

the matched invariant envelope is a generalized fixed point in the envelope

phase space. This is an important observation having implications for parti-

cle motion within a slice.

2.2 Space-Charge Dynamics within a Slice

The emittance oscillations examined above are the result of the fact that the

line charge density, λ (ζ), varies from slice to slice, which in turn causes

the trace space trajectories of the slices to differ. Emittance compensation

can occur because, while the amplitude of trace space oscillations varies be-

tween slices, the oscillation frequency, to first order, does not. Above, the

slice trajectories were found by assuming a cold beam limit — where each

slice makes a perfect line in trace space, and has zero emittance — and that

the transverse space-charge forces were linear, and therefore that the slice

density was independent of radial position. In considering the rms motion

of the slices, we relaxed the requirement of constant slice density by consid-

ering only the effect of the linear component of the transverse space-charge

force.

If the slice density is not uniform, the envelope analysis given above is still

84
valid, but not complete, as it ignores the effect non-linear space-charge forces

have within a single slice. This section considers the role of these forces in

the evolution of a single slice. In particular, it will be shown that when a

nonuniform density slice is matched (in an rms sense) to an equilibrium,

or a generalized equilibrium, such an the invariant envelope, it undergoes

wave-breaking in the transverse phase space — where the previously approx-

imately single-valued distribution of transverse momenta, fpx (x) becomes

multiple valued — thus causing an irreversible emittance growth.

This emittance growth mechanism has been studied extensively in the

field of heavy-ion fusion in the context of Brillouin flow in coasting, solenoid-

focused beams. It is well understood from the viewpoint of microscopic

phase space dynamics of coasting beams, as studied by Anderson [67], and

alternatively as the conversion of so-called nonlinear field energy to thermal

energy, and thus emittance [67, 68, 69]. This irreversible emittance growth

has also been associated in O’Shea’s analysis with the increase in the beam

entropy [70]. The treatment of this topic given below is concerned primarily

with purely transverse microscopic dynamics associated with beams under-

going reversible and irreversible emittance growth, and therefore, is strongly

connected to Anderson’s study of the microscopic dynamics in space-charge

dominated ion beams. Conversely, in the cases of present interest we are not

concerned with long-range irreversible behavior, such as halo formation, or

equipartitioning of energy between phase planes, and so we do not find it

necessary to use thermodynamics-based tools more familiar to the ion beam

community.

Therefore, this section is concerned with the self-consistent phase space

85
dynamics of beam slices as they evolve under the influences of space-charge

and external forces. These dynamics are analytically studied to determine

the conditions under which phase space wave-breaking occurs, for coasting

beams in slab-symmetric, as well as cylindrically symmetric, geometries. The

slab-symmetric case is included mainly to allow use of exact and physically

transparent results, which illustrate the mechanisms involved in phase space

wave-breaking. In practice, one is nearly always concerned with cylindrically

symmetric beams, and so this analysis is extended in this case to include

acceleration in an rf structure. Because the analysis of this system is not

tractable after wave-breaking has occurred, computational simulations are

also employed to further the investigation of cylindrically symmetric beam

physics in both the coasting and accelerating cases. The results of this anal-

ysis show that, in order to compensate the beam emittance within a slice,

in the presence of significant nonlinearities of the space-charge fields, one

must avoid matching of the beam to the generalized equilibria, and that the

optimal transport of a space-charge dominated beam is typically not close to

such equilibria.

2.2.1 Slab Beams

As can be seen from Sections 2.1.3 and 2.1.4, the self-consistent collective

motion of particle beams in cylindrical symmetry is complicated somewhat

by the need to approximate the solutions to the relevant differential equa-

tions. Because of this, it is most instructive to begin the analysis using a

Cartesian, or slab-symmetric (sheet) beam, following the general methods

86
introduced by Anderson in Ref. [67].

We start this discussion by examining a freely expanding (unfocused) lam-

inar beam, with initial (z = 0) density profile, infinite in the y and z dimen-

sions, and propagating in the +z direction. In this case, we may write the

beam density as

Σb
nb (x0 ) = f (x0 ) (with f (0) = 1), (2.68)
a0

where Σb is the beam charge per unit (slab) area and a0 = Σb /nb0 is the

effective initial beam width. The case of free expansion can be considered to

ve the most nonequilibrium scenario possible. It can also be thought of as

forming one portion of propagation under periodic application of thin lenses

separated by drifts of external force-free regions.

The equations of motion for the electron position for the free-expansion

scenario, under laminar flow conditions, are derived in a manner similar to

Equation (2.12), and are found to be

 x0



x (z) = k2p0 F (x0 ) , F (x0 ) = f (x) dx = constant, (2.69)
0

where the initial plasma wave number of the slice is defined by replacing

nb (ζ, z) with nb0 in Equation (2.11). If laminarity is obeyed, the integral

F (x0 ) is constant and these equations have solutions dependent only on

initial conditions,

2
kp0 z
x (x0 , z) = x0 + F (x0 ) , (2.70)
2

again assuming that x0
= 0 for all particles in the slice. The density distribu-

tion is also a simple function of its initial state, as conservation of probability

87
gives f (x (x0 ) , z) dx = f (x0 ) dx0 , or

f (x0 ) f (x0 )
f (x (x0 ) , z) = dx(x0 )
= 2 . (2.71)
(kp0 z )
dx0 1+ 2
f (x0 )

In the freely expanding case, the density distribution becomes more uniform


as it expands over many plasma radians kp z  1 ,

f (x0 ) 2
f (x (x0 ) , z) = 2 →
2 . (2.72)
(kp0 z ) k z
1+ 2
f (x0 ) p0

This observation is critical, as it implies that the transport is “more lin-

ear,” since the space-charge defocusing for a uniform beam becomes approx-

imately linearly dependent on offset,

x
x

(z) = k2p0 F (x0 ) ≈ . (2.73)
2z2

This will in turn imply that the phase space wave-breaking effects which lead

to irreversible emittance growth are mitigated, since the angle that a particle

attains becomes more linearly correlated with position,

x
k2p0 zF (x0 ) 2
= 1 2 → . (2.74)
x x0 + 2 kp0 z2 F (x0 ) z

If the phase space distribution lies along a straight line, the emittance van-

ishes, so Equation (2.74) indicates a desirable trend.

An example of this increased distribution uniformity is shown in Fig-

ure 2.10, where a beam with initial parabolic profile

 2
x0
f (x0 ) = 1 − (2.75)
a

has freely expanded for a distance kp0 z = 4. The profile has become notice-

ably flattened during this expansion.

88
 1.2

0.8

f (0) 0.6
f

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2

x
xmax

Figure 2.10: Initially parabolic slab beam distribution (solid line) mapped to
a more uniform (normalized) distribution (dashed line) after a drift length
x
kp0 z = 4. Distribution shown as a function of relative offset position xmax
.

It is instructive at this point to calculate the emittance evolution associ-

ated with the freely expanding beam. In order to do so, we consider a number

of possible forms of the distribution: Gaussian, parabolic, and uniform (“flat-

top”). The single particle equations of motion and the condition of laminar

flow allow the calculation of the second moments of the distribution and

consequently the rms emittance. Laminar flow implies that

nb (x, z) dx = nb (x0 ) dx0 . (2.76)

Thus, the second moments can be simply calculated by integrating over the

initial particle positions using the initial position x0 as the integration vari-

89
able. For example, σx2 is given by
∞
x  =
2
x 2 (x0 , z) nb (x0 ) dx0 , (2.77)
−∞

or explicitly,
∞  2
1 2 2
x  =
2
x0 + kp0 z F (x0 ) nb0 (x0 ) dx0 , (2.78)
−∞ 2

which simplifies to

1 4 4 2
x 2  = x02  + k2p0 z2 x0 F  + k z F . (2.79)
4 p0

Similarly, the other moments are

x
2  = k4p0 z2 F 2 , (2.80)

and
1 4 3 2
xx
 = k2p0 zx0 F  + k z F . (2.81)
2 p0
From this, the emittance is given by

εx = k2p0 z x02 F 2  − x0 F 2 , (2.82)

or

εx = αk2p0 σx0 z. (2.83)

The α in Equation (2.83) is the quantity under the radical in Equation (2.82),

normalized by the initial rms width, σx0 , and is therefore, a form factor that

depends only on the initial beam distribution type. The values of α are sum-

marized in Table 2.1 for a few commonly considered distribution types. It

is clear from Equation (2.82) that the emittance remains zero if F (x0 ) = x0 ,

which only occurs if the distribution is uniform. The quantity under the rad-

ical in Equation (2.82), and therefore α, are measures of the nonuniformity

90
 Profile α

π−3
Gaussian 3π

2
Parabolic 3675

Flat-top 0

Table 2.1: Values of the form factor α for various slab-symmetric initial beam
distribution types

of the initial distribution. Equivalently, α can be viewed as a measure of the

nonlinearity of the space-charge fields of the beam, since, according to Equa-

tion (2.69), the space-charge force is proportional to F (x0 ). This notion will

be useful in the conceptual understanding of emittance oscillations that will

be found to occur in beam slices where an external focusing force is applied.

Note that in the case of free expansion the emittance grows linearly with

distance from the launching point, but has no dependence on initial beam

size, as k2p0 σ0 ∝ Σb . While this linear growth is a worrisome phenomenon,

it turns out not to be valid for cylindrically symmetric beams. In cylindrical

beams, the emittance growth is reversed after a time during expansion, and

after application of a thin lens, a nearly perfect oscillation of this nonlinear

space-charge force-induced emittance can be made to occur. This compensa-

tion of the nonlinearity-derived emittance will be discussed in the following

sections.

Wave-breaking occurs in phase space when the value of x (z) somewhere

in the distribution becomes independent of x0 , and the transverse momen-

tum distribution becomes a multiple valued function of transverse offset.

91
According to Equation (2.69), this condition (dx/dx0 = 0) also implies that

the density becomes singular at these points. Note that there is no wave-

breaking for the free expansion slab-symmetric case, as

2
dx kp0 z
= 1+ f (x0 ) > 1 > 0. (2.84)
dx0 2

This will change when we introduce focusing, but one conclusion remains

from this analysis: one must allow the beam to stay far from equilibrium in

order to avoid the most serious consequences of wave-breaking.

There are two ways to proceed from this point. One is to introduce thin

lenses to produce a periodic transport system with an rms matched (in the

sense that the envelope has the same periodicity and symmetry as the applied

focusing forces) beam. The other is to introduce a uniform strength focusing

channel (akin to the solenoid in cylindrically symmetric systems), but to allow

a mismatch between the beam and the channel. In the interest of analytical

simplicity, the latter case is examined first.

In a system with uniform strength focusing, Equation (2.69) becomes

x

(x) + k2β x (z) = k2p0 F (x0 ) , (2.85)

where we use the betatron wave number kβ , associated with free oscillations

under the influence of the focusing gradient. The equilibrium solution for a

given initial particle position is simply

k2p0
xeq (x0 ) = F (x0 ) . (2.86)
k2β

This equilibrium can be made consistent for all particles, in the sense that no

particles will move after the distribution is launched, if F (x0 ) = x0 and k2p0 =

92
k2β . If any initial distribution other than a uniform one is employed, there will

be subsequent motion and associated rearrangement of the distribution. In

this more general case, we may write the solution to Equation (2.85) as

 

x (x0 , z) = xeq (x0 ) + x0 − xeq (x0 ) cos kβ z . (2.87)

The wave-breaking condition associated with this motion is

 
dx k2p0 k2p0

= 2 f (x0 ) + 1 − 2 f (x0 ) cos kβ z = 0, (2.88)
dx0 kβ kβ

or


k2β cos kβ z
f (x0 ) = − 2
. (2.89)
2kp0 sin2 kβ z/2

It can been seen from this that wave-breaking always occurs for a suffi-

ciently small value of f (x0 ), i.e., portions of the beam found in a long con-

tinuous tail, assuming a monotonically decreasing function f (x0 ). Quan-

titatively, Equation (2.89) states that wave-breaking eventually occurs for

all f (x0 ) < k2β /2k2p0 , with the most interior value of x0 undergoing wave-

breaking at kβ z = π (for distributions which smoothly approach zero, wave-

breaking begins in these tails at kβ z = π /2). It is also apparent that wave-

breaking can be avoided by a combination of removal of the tails, so that

f (x0 ) discontinuously goes to zero at a hard-edge beam boundary, and

by making the ratio k2β /k2p0 small. When this ratio is near unity, the beam

is closely “matched” to the external focusing, and when the ratio is much

smaller than unity the beam is mismatched, with the focusing being too weak

to control the beam distribution at its launch size.

An alternative way of understanding wave-breaking is to note that the

equilibrium beam size xeq , associated with the initial wave-breaking position

93
 0.20

0.15
Distribution
0.10
Trace space
0.050
x
rotation
0.0

-0.050
Fixed point
-0.10

-0.15

-0.20
0.0 0.20 0.40 0.60 0.80 1.0 1.2

Figure 2.11: Trace space picture of a slab symmetric beam at wave-breaking

onset (kβ z = π /2), for the case of k2β /k2p0 = 2/3, showing two fixed points
with opposing direction of rotation.

is a fixed point of the oscillation. Also, we know that the origin in trace space

is also a fixed point, with an opposing sense of phase space rotation about it.

The existence of two such fixed points guarantees that the trace space will

filament after wave-breaking and the emittance will grow irreversibly. The

trace space picture of this system is displayed in Figure 2.11.

Thus we deduce that a mismatched beam is more likely to preserve its

laminar flow, under mismatched conditions, which is an extension and deep-

ening of what we have learned from the case of free expansion. To em-

phasize this point, Figure 2.12 shows a plot of normalized beam density at

the maximal wave-breaking point, kβ z = π , for a cutoff (at the 25% intensity

level) parabolic distribution in nearly matched (k2β /k2p0 = 4/3) and highly mis-

94
 5.0

4.0
mismatched
matched

f /f (0)
3.0

2.0

1.0

0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2

x/xmax

Figure 2.12: Normalized beam density f /f (0) for a beam with initially
parabolic slab distribution (cutoff at 0.25 normalized density) at kβ z = π
for distributions in nearly matched (k2β /k2p0 = 4/3) and highly mismatched
(k2β /k2p0 = 1/3) cases. Offset x is normalized to its maximum value in the
distribution.

matched (k2β/k2p0 = 1/3) cases. The nearly matched case barely evades wave-

breaking, and displays a very large density spike at the beam edge, while the

highly mismatched beam easily maintains laminarity, giving a much smaller

density spike.

In order to calculate the emittance evolution in the case of the slab beam

in a focusing channel, we follow the same procedure as in the drifting beam

case up to the point of wave-breaking, where laminar flow strictly ends, and

out present analysis breaks down. Assuming a cold beam initially at a waist

(x0
= 0), the emittance evolution is found to be

k2p0 

εx = α σ0 sin kβ z  , (2.90)
kβ

where, again, α is a constant depending on the form of the initial distribution

95
and the factor k2p0 σ0 ∝ Σb has no dependence on initial beam size. Note from

this that the predicted maximum emittance occurs at kβ z = π /2, as with the

correlated inter-slice emittance studied in emittance compensation. It should

also be emphasized that this is the same longitudinal position in which the

initial wave-breaking occurs for a distribution with a continuous tail.

This emittance oscillation — which occurs at twice the plasma frequency

for a matched beam, just as the oscillation associated with emittance com-

pensation — has been analyzed in the context of intense ion beams [69],

as an oscillation in the energy of the system, between kinetic energy of the

beam particles and potential energy stored in the space-charge field. In three

dimensions, the field energy Wsc within a volume V is given by


0
Wsc = 2 dV .
E (2.91)
2
V

As indicated by Equation (2.82), and rigorously proven [68], the field energy

Wsc is minimized when the beam’s distribution is uniform. If the distribution

is nonuniform, the configuration will have a greater Wsc , and the difference

Wsc −Wmin , called the nonlinear, or free self-field energy, is a form of potential

energy that can be converted into kinetic energy of the beam particles, and

thus emittance growth. This energy oscillation also explains the tendency of

the beam profile to become more uniform, as shown in Figures 2.10 and 2.12,

as the emittance increases, and field energy is converted to kinetic energy. As

the emittance decreases, we expect the distribution to become less uniform,

ultimately returning to its initial shape when the emittance goes to zero.

To observe this effect, self-consistent simulations must be used, as the beam

undergoes wave-breaking at the emittance maximum, and the analysis relying

96
on laminar flow breaks down. These simulations are shown below for the case

of cylindrically symmetric beams.

2.2.2 Cylindrically Symmetric Beams

To extend this analysis to beams with cylindrical symmetry, we proceed in

a manner similar to that of the previous section, and find that many of the

previous results in this chapter apply here as well. To begin with, the beam

density in an axisymmetric system can be written in a form similar to Equa-

tion (2.1),

nb (r , z) = λb f (r , z) , (2.92)

where λb = I/ev is the beam’s axial charge density, and the ζ dependence of

λ is ignored, since we are considering the dynamics within a single slice. The

analysis leading to Equation (2.12) applies here, but in this case in amount

of beam charge within a given initial radius r0 must be considered more

carefully, as it depends on the initial distribution f (r0 ). The electromagnetic

force on a particle in such a distribution, assuming laminar flow, is

 r0
2e 2eλ (r0 )
Fr (r , z) = 2 2π nb (r̃ , z) r̃ dr̃ ≡ , (2.93)
γ r 0 γ 2r

where λ (r0 ) is defined as the total charge enclosed by an initial point r0 (z0 ),

according to
 r0
λ (r0 ) = 2π nb (r̃ , z0 ) r̃ dr̃ . (2.94)
0

Notice that here, λ (r0 ) plays the role of F (x0 ) in the slab symmetric case,

and it is a constant of the motion for laminar flow.

The paraxial equation of motion for a particle with no canonical angular

97
momentum experiencing both a solenoidal restoring force and the repulsive

space charge force corresponding to Equation (2.93) is

2re λ (r0 )
r

(z) + k2β r (z) = . (2.95)
β2 γ 3 r

This equation, like Equations (2.12) and (2.54), is a nonlinear equation not

amenable to exact solution in general. We can begin an approximate analysis,

however, by defining an equilibrium radius corresponding to each value of

r0 , 

 2re λ (r0 ) k̄p (r0 )
req (r0 ) =  2 2 3 ≡ r0 √ . (2.96)
kβ β γ 2kβ

Here we have introduced an average beam plasma frequency

4π re n̄b (r0 ) 2re λ (r0 )

k̄2p (r0 ) = 2 3
= 2 2 3 , (2.97)
β γ r0 β γ

which corresponds to the mean enclosed initial density at r0 .

Equation (2.95) can be linearized about the equilibria given by Equation (2.96)

to obtain

δr

+ 2k2β δr = 0, (2.98)

where δr = r0 − req . This equation yields a familiar form of solution, for a

distribution beginning with no radial momentum (or angular momentum in

the beam’s Larmor frame)

   
r (r0 , z) = req (r0 ) + r0 − req (r0 ) cos 2kβ z . (2.99)

The wave-breaking condition is again given by

 
dr dreq dreq  
= + 1− cos 2kβ z = 0, (2.100)
dr0 dr0 dr0

98
or √ 
dreq cos 2kβ z
=−  √ . (2.101)
dr0 2 sin2 kβ z/ 2
The quantity on the left-hand side of Equation (2.101) can be written as

 k2p (r0 )
dreq 1 2re 1 dλ
=  2 2 3 =√ , (2.102)
dr0 2 kβ β γ λ (r0 ) dr0 2kβ k̄p (r0 )

where the local value of the initial beam plasma frequency is given by

4π re nb (r0 ) 2re λ (r0 )

k2p (r0 ) = 2 3
= 2 2 3 . (2.103)
β γ r0 β γ

As an illustrative example, consider the wave-breaking condition for the

case of an initially Gaussian beam, where

r02
nb (r0 ) = nb0 exp − . (2.104)
2σr2

In this case,
 r0

r2
λ (r0 ) = 2π nb0 r exp − dr
0 2σr2


2 r02
= 2π nb0 σr 1 − exp −
2σr2

= 2π σr2 [nb0 − nb (r0 )] , (2.105)

and the wave-breaking condition can be written as

  √ 
kp0 r0 exp −r0
2
/2σ 2
r cos 2k β z
  =−  √ . (2.106)
kβ σr sin2
k z/ 2
1 − exp −r0 /2σr
2 2 β

For wave-breaking to be avoided, the left-hand side of this equation must be

greater than unity,

 
kp0 r0 exp −r02 /2σr2 kp0
kβ σr   ≡ k g (r0 ) > 1. (2.107)
β
1 − exp −r02 /2σr2

99
 1.2

1.0
f
g

g/g (0)
0.80

0.60

f /f (0) ,
0.40

0.20

0.0
0.0 0.50 1.0 1.5 2.0 2.5 3.0 3.5

r /σr

Figure 2.13: A comparison of the function g (r0 ) with Gaussian f (r0 ).

The function g (r0 ) is shown in Figure 2.13 with f (r0 ) also displayed for

comparison. It can be seen that g (r0 ) approximately follows the density, and

thus the threshold for wave-breaking is estimated as

kp0
f (r0 )  1. (2.108)
kβ

This is in contrast to the equivalent condition found in the slab beam case,

k2p0
f (x0 )  1, (2.109)
2k2β

which has the stronger quadratic dependence on the mismatch parameter

kp0 /kβ .

As the linear dynamics of the axisymmetric beam have been seen to be

formally quite similar to those of the slab beam, it is not surprising that the

emittance evolution is similar as well. Given the same initial conditions as

assumed in the slab case, we find the emittance to be of the same form as

100
 Profile α

Gaussian 0.141

Parabolic 0.065

Flat-top 0

Table 2.2: Values of the form factor α for various cylindrically symmetric
initial beam distribution types

well,
  
 
εr = ασ02 kp0 sin 2kβ z  . (2.110)

Here, α is again a form factor, defined as in the previous section. The nu-

merical values of α found for the cylindrically symmetric case are shown in

Table 2.2. It will be shown below, that Equation (2.110) provides a very ac-

curate description of the emittance evolution up until wave-breaking. Note

that the emittance in Equation (2.110) is in fact linearly dependent on σ0 , as

kp0 ∝ σ0−1 .

2.2.3 Simulation of Coasting Cylindrical Beams

The analytical treatments of intra-slice transverse space-charge detailed above

are limited to the laminar flow regime and, in the case of cylindrical beams,

are only approximate. They do, however, predict where wave-breaking will

occur, and that it can be minimized or avoided by mismatching the beam-

focusing channel system. In order to test these predictions and examine the

behavior of the beam after wave-breaking, we use self-consistent simulations,

using a one-dimensional time dependent code called norse (described in Ap-

101
pendix A.2), that follow the evolution of the beam using the space-charge

force of Equation (2.93).

We found in the case of the slab beam expanding under its space-charge

force that there was no wave-breaking for any type of distribution. Equa-

tion (2.93) indicates that this is not true for a freely expanding cylindrical

beam if the initial distribution function falls off, so that the integral of the

charge density does not increase proportionally with r . In this case we ex-

pect wave-breaking and look to simulations for understanding of the beam

behavior after wave-breaking occurs.

In the simulations, the emittance evolution of the freely expanding beam

shows the effects of wave-breaking. As in the focusing channel, the emit-

tance increases to a maximum at z = λp0 /4 = π /2kp0 where wave-breaking

occurs. While the beam continues to expand, the particles in the vicinity of

the initial wave-breaking point (where the maximum outward force is found)

effectively rotate about this outward-moving point. This rotation causes the

tail particles to “tuck under” in phase space in a distance a bit longer than

the initial plasma half-wavelength (the plasma frequency is not constant, but

decreases as the beam expands), as would be expected, and the emittance

decreases during this initial rotation. The emittance growth is not perfectly

compensated by this nonlinear effect, however, and the emittance reaches

a local minimum. After that, ε becomes simply proportional to σ as the

beam continues to expand. Examination of the beam’s phase space evolu-

tion, shown in Figures 2.14 and 2.15, illustrate this process.

Note from Figure 2.15b, that this tuck under effect on the emittance occurs

102
 20 1 0

εr [Arbitrary Units]
15 7.5

σ /σ0 1 0 5

5.0 2.5

0.0 0
0.0 0.25 0.50 0.75 1.0

kp0 z/2π

Figure 2.14: Results of a simulation of the free expansion of an initially

Gaussian beam. The beam size (solid line) increases monotonically while the
emittance (dashed line) has a local maximum and minimum.

0.10 0.15

0.08 0.12
r
[radians]

r
[radians]

0.06 0.09

0.04 0.06

0.02 0.03
(a) (b)

0 .0 0.0
0.0 1.0 2.0 3.0 4.0 5.0 0.0 5.0 10 15

r [mm] r [mm]

Figure 2.15: Trace space plots of a freely expanding, initially Gaussian beam
at the initial emittance (a) maximum and (b) minimum.

103
only after the rms beam size has grown substantially (recall that kp0 z > π ,

and the beam has had a large distance in which to expand), as the emittance

minimum occurs when σ /σ0 ≈ 8.5.

While the drifting beam is instructive, we are interested in beam transport

involving focusing elements. We proceed again as before, by examining two

cases: periodic thin lens focusing separated by drifts, and a focusing channel.

In the case of thin lens focusing we can directly apply the result of the drifting

beam. Simulations show that, for a given transport length, fewer lenses and

larger beam size oscillations will produce a lower emittance at the end of the

transport line, provided that the beam makes an integer number of envelope

(plasma) oscillations. Figures 2.16 and 2.17 show two simulations of a beam

with the same initial conditions and transported through the same length

of drift. In the first there is one thin lens applied when σ /σ0 = 8.5. In the

second, in order to approximate a beam which is more closely matched to a

uniform focusing channel, a lens is applied each time the beam size doubles

its initial value. It is clear from the graphs that, when the beam is allowed

to expand enough to take advantage of the tuck under effect observed in the

drifting beam above, much of the emittance growth can be reversed when

the beam is focused back down. In the case where the beam size oscillations

are kept smaller we see that the emittance oscillates about its peak value but

never drops to as low a level as in the first case.

The striking performance of the scheme shown in Figure 2.16 for minimiz-

ing the emittance at the envelope minimum — in other words, compensation

of the nonlinear field-derived emittance — is understandable in a number of

different ways. If the dynamics being described were only the linear slice dy-

104
 10 6

εr [Arbitrary Units]
7.5 4.5

σ /σ0
5.0 3

2.5 1.5

0.0 0
0.00 0.29 0.58 0.87 1.16 1.45

kp0 z/2π

Figure 2.16: Evolution of beam size and emittance in a simulation with thin
lens focusing applied at the point of the initial emittance minimum. The lens
strength is chosen to reverse the envelope angle.

2.0 6

εr [Arbitrary Units]
1.5 4.5
σ /σ0

1.0 3

0.50 1.5

0.0 0
0.0 0.3 0.6 0.9 1.2 1.5

kp0 z/2π

Figure 2.17: Evolution of beam size and emittance in a simulation with thin
lens focusing applied at the points of beam envelope doubling and lens
strength chosen to reverse the envelope angle. The simulation is followed
for the same distance as in Figure 2.16.

105
namics, Figures 2.7 and 2.8 illustrate that the emittance performance would

be qualitatively the same in Figures 2.16 and 2.17. They are not, however,

and this is because of the strong wave-breaking induced in the intra-slice

dynamics by the beam being too close to equilibrium. In other words, the

existence of the off-origin moving “fixed point” in trace space gives rise to

wave-breaking, trace space filamentation, and associated irreversible emit-

tance growth. O’Shea has identified irreversible emittance growth of this

type with an increase in the entropy [70] which, we note, is also equivalent

to loss of order or information in the system. In the case of Figure 2.16, the

emittance increase due to field nonlinearities is reversed (compensated) and

the information about the beam’s initial state is recovered. This idea of infor-

mation recovery is consistent with the concepts of energy exchange between

the particles and the space-charge fields, described above. The phenomenon

is illustrated in Figure 2.18, which shows the beam distribution in r at three

points in the propagation of Figure 2.16 — the initial and final states, as well

as the thin lens position. It can be seen that, by judicious choice of focus-

ing, the final beam distribution reproduces the initial distribution remarkably

well (information is retained during the beam oscillation), considering how

distorted it becomes in intermediate points in the propagation.

It is useful at this point to make a connection between the photoinjector

terminology employed in this thesis, based on collective space-charge forces

and stated in terms of plasma frequencies, and the terminology of the ion-

beam community, which emphasizes the optical properties of the periodic

focusing system used. Periodic focusing systems are parameterized, in the

limit of no collective forces (emittance dominated beams), by the betatron

106
 700

[macroparticles/mm2 ]
600

500

Density
400

300

200
(a)
100

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2

r [mm]

16
[macroparticles/mm2 ]

14

12

10
Density

2 (b)
0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0

r [mm]

700
[macroparticles/mm2 ]

600

500
Density

400

300

200

(c)
100

0
0.0 0.2 0.4 0.6 0.8 1. 0 1 .2

r [mm]

Figure 2.18: Evolution of the beam distribution during the simulation shown
in Figure 2.16 at the (a) beginning, (b) focusing lens (midpoint), and (c) end-
point (emittance minimum).

107
phase advance per period µ. The case shown in Figure 2.16, where the beam

expands dramatically between lenses, is one in which µ is undefined, i.e., the

transport is unstable with no space-charge focusing present, while the space-

charge depressed phase advance is near vanishing. This situation is known

to be prone to envelope instability [3], however, as well as halo formation.

For long periodic transport systems, these properties would be highly unde-

sirable, but for the one or two oscillation systems typified by photoinjectors

(where the space-charge tune depression is removed by acceleration) these

effects do not have time to assert themselves. On the other hand, in the case

shown in Figure 2.17, where a large amount of irreversible emittance growth

is observed due to wave-breaking effects, µ is small compared to unity, and

the envelope motion is stable. Thus we are led to an ironic conclusion — that

the nearly matched, small µ transport is stable in an envelope sense, but very

damaging from the viewpoint of emittance compensation.

It is natural to consider the limit suggested by the simulation of Fig-

ure 2.17, in which the beam envelope does not vary at all — the case of a

beam matched in the rms sense to a uniform solenoidal focusing channel. We

can also compare these simulations with the prediction of Equation (2.110),

at least until the onset of wave-breaking. The emittance evolution found by

simulation of an initially parabolic beam rms matched to a focusing chan-

nel along with the emittance predicted by Equation (2.110) is shown in Fig-

ure 2.19. Note that the emittance again follows the same pattern shown above

in that it increases rapidly in a quarter of a plasma oscillation to a maximum

[67]. Since wave-breaking does not occur until this maximum is reached, the

excellent agreement between theory and simulation up to that point is not

108
 1.2

1.0

εr ,n [mm mrad]
0.8

0.6

0.4
Simulation
Theory
0.2

0.0
0.0 5.0 102 1.0 103 1.5 103 2.0 103 2.5 103

z [mm]

Figure 2.19: Evolution of the emittance for a beam rms matched to a uniform
focusing channel from simulation and analytical prediction (Equation (2.110)).

surprising. We will encounter a similar type of emittance behavior in accel-

erating systems in the following section.

2.2.4 Accelerating Cylindrical Beams

For a beam accelerating under the influence of radio frequency fields, the

paraxial equation of motion for a particle in a laminar condition now con-

tains terms, as was found for the envelope equation, arising from adiabatic

damping and ponderomotive (alternating transverse gradient) forces,

2
γ
η γ
2re λ (r0 )
r

(z) + r
(z) + r (z) = , (2.111)
γ (z) 8 γ (z) γ (z)3 r (z)

which is again a nonlinear equation without an analytical solution. In this sys-

tem, there is also an equilibrium-like particular solution to Equation (2.111),

109
which is analogous to the invariant envelope above, corresponding to each

value of r0 .

As before, the procedure here is to find an analytical formula for the

emittance of a “matched” beam. In the case of an accelerating beam, the

term matched is meant in the sense that the rms size of the beam follows the

invariant envelope. This situation is slightly different from that of coasting

beams because we are required to reference σ
to the non-stationary partic-

ular solution

4 re λ (r0 )
rp (r0 , z) =

γ (2 + η) γ (z)

k̄p (r0 ) 1 γ0
≡ r0 . (2.112)
kβ 2 + η γ (z)
√
In Equation (2.112) we have identified kβ = γ
/ 8γ, and can see that the

particular solution is again proportional to the initial ratio k̄p (r0 ) /kβ . We can

again proceed to linearize Equation (2.111) about these particular solutions

to obtain


2


γ

1+η γ

δr + δr + δr = 0, (2.113)
γ 4 γ
where δr = r −rp . This equation has a general form of solution similar to that

given by Equations (2.66) and (2.67). Therefore, we can solve Equation (2.111)

to find the single particle motion, yielding



  1+η γ0
r (r0 , z) = rp (r0 , z) + r0 − rp0 (r0 ) cos ln
2 γ

 (2.114)
1   1+η γ0
+ r0 − rp0 (r0 ) sin ln ,
1+η 2 γ

where the integration constants are chosen so that

1 γ

r0
= r0 . (2.115)
2γ

110
The wave-breaking condition is now given by
√  
1+η γ
drp cos 2
ln γ0
=− √   . (2.116)
dr0 2 1+η γ
2 sin 2
ln γ0

The right-hand side of Equation (2.116) can be recast to give


drp 4π r0 nb re
=
dr0 γ (2 + η) λ (r0 ) γ (z)
2

kp (r0 ) η
=
2kβ γ
2+η
√  
1+η γ
cos 2
ln γ0
=− √   , (2.117)
1+η γ0
2 sin2 2 ln γ

and we see that wave-breaking is again averted by cutting the tails off of the

distribution.

To proceed in the analysis, we can follow the same procedure as before,

and use the laminarity condition to integrate over the initial beam distri-

bution to determine its second moments and emittance. Performing this

calculation for a beam rms matched to the invariant envelope gives

 

 1+η 
4αre λb  γ0 
εr ,g =   sin ln . (2.118)
π (1 + η)γ γ0 γ 

3 2 γ 

The subscript indicating that the emittance is the geometric (or un-normalized)

measure is included here to make this point clear, since, because the beam

is accelerating, εr ,g is a function of γ (z). In Equation (2.118), as before, α is

a constant depending on the initial beam distribution, with values listed in

Table 2.3.

The expression for the emittance evolution given in Equation (2.118) is

valid (in the linear approximation |δr |  rp ) up to the wave-breaking point.

111
 Profile α

Gaussian 0.170

Parabolic 0.056

Flat-top 0

Table 2.3: Values of the form factor α for various cylindrically symmetric
initial beam distribution types, accelerating case.

The details of wave-breaking in the accelerating beam system are discussed

below. Note that the emittance for this case is inversely dependent on the

acceleration gradient γ
, and proportional to the beam current. These de-

pendences are due primarily to the setting of the beam size to be near the

invariant envelope.

Simulations are again useful in the accelerating beam case to check the an-

alytical predictions given above, and to investigate the behavior of the beam

slice after wave-breaking. Figure 2.20 shows the simulation of the normalized

emittance evolution in the case of an initially parabolic profile accelerating

beam matched to the invariant envelope along with the prediction of Equa-

tion (2.118). Again we see that the normalized emittance rapidly increases

to a local maximum. We also see from the figure that the analytical formula

for the emittance agrees well with the simulation up to the emittance max-

imum. However, because Equation (2.113) is the linearized equation of mo-

tion, and |δr | has constant amplitude, while rp ∝ γ −1/2 ∝ z−1/2 decreases,

the agreement between theory and simulation is not as striking as with the

coasting beam. Also, we see that theory and simulation do not agree after

112
 1.4

1.2

εr ,n [mm mrad] 0.8

0.6

0.4

Theory
0.2 Simulation

0
0 1000 2000 3000 4000 5000 6000

z [mm]

Figure 2.20: Emittance evolution of an initially parabolic beam matched to

the invariant envelope with a 60 MV/m peak acceleration field gradient. (The
beam and accelerator parameters are the same as those for the Neptune
PWT.) The dashed line is the simulation result and the solid line is produced
by Equation (2.118).

the emittance maximum. This is in keeping with the coasting beam results,

as the beam undergoes wave-breaking near the emittance maximum and the

assumption of laminarity used in Equations (2.111)–(2.118) is no longer true.

This wave-breaking is easily seen in the beam trace space at the peak emit-

tance shown in Figure 2.21.

Observe in Figure 2.21 that the emittance does not change significantly

shortly after the emittance maximum. Since the transverse plasma frequency

of the beam decreases as γ −3/2 , the acceleration process essentially stops

113
 0 100

-1 10-2

r
[radians]

-2 10-2

-3 10-2

-4 10-2
0 0.1 0.2 0.3 0.4 0.5 0.6

r [mm]

Figure 2.21: Trace space of an initially parabolic beam slice at the maximum
emittance point in the accelerating beam simulation. wave-breaking has just
occurred.

the plasma oscillations and the beam becomes emittance dominated. The

initial emittance growth caused by space-charge field nonlinearities then, is

“frozen in,” and the beam has a finite irreversible emittance. We can use

Equation (2.118) to estimate the final emittance of the beam and, therefore,

its size in the emittance dominated limit. To do this, we start by finding the

114
position of the emittance maximum,

dεr2,n
=0
dz

2 

16α2 re λb 1+η γ0
=− sin ln (2.119)
π (1 + η) γ0 γ
γ 2 2 γ
! 
  
"
1+η γ0 1+η γ0
× sin ln + 1 + η cos ln ,
2 γ 2 γ
or

 
1+η γ0
tan ln = − 1 + η. (2.120)
2 γ
Equation (2.120) yields the position of the emittance maximum,
 
γ0 1
zεmax =
 %√
 & − 1 , (2.121)
γ −1
exp 2 tan − 1+η

and the maximum emittance at this point is


 
4αre λb sin tan−1 − 1 + η
εn,max = %√
 & . (2.122)
π (1 + η)γ0 γ
−1
exp 2 tan 1+η

The final beam size in the simulations is estimated by ignoring the space-

charge term in Equation (2.52), and assuming a steady state solution based on

a constant normalized emittance equal to the maximum predicted by Equa-

tion (2.122),

1/4 
8 εn,max
σmin = . (2.123)
η γ

A comparison between the final rms beam size achieved in simulation and

the prediction of Equation (2.123) for the simulation case of Figure 2.20 is

shown in Figure 2.22. The agreement is remarkably good in the asymptotic

region, where the simulated beam size approaches a constant value very close

to that predicted from the analytical result of Equation (2.123). Thus, on

can determine the final beam characteristics simply by knowing the degree

115
 1

Theory
Emittance Limit
0.8 Simulation
IE

σr [mm] 0.6

0.4

0.2

0
0 1000 2000 3000 4000 5000

z [mm]

Figure 2.22: The beam envelope evolution for the same simulation as Fig-
ure 2.20. Here the beam size follows the invariant envelope initially, but
levels off as it approaches the limit predicted by Equation (2.123).

of nonuniformity of the distribution (which is parameterized by α) at the

beginning of acceleration with transverse matching to the invariant envelope.

As an example of the potential final emittance, we take the nominal LCLS

photoinjector design parameters [12], in which a 100 Amp beam is emitted

in a high gradient rf gun, accelerated to γ0 ≈ 14, and then focused into a

matched invariant envelope at the beginning of a high gradient linac. For

a standard SLAC S-band traveling wave (η  0.3) linac (average accelerating

gradient of 17 MeV/m), one obtains an asymptotic emittance of εn,max = 6.5α

mm mrad. Even though a roughly uniform beam is planned to be launched

116
at the cathode in this device, it will be nonuniform at the injection to the

linac due to nonlinearities in the space-charge forces at very low velocities,

as well as imperfections in the drive laser spatiotemporal profile. To see the

potential effects of such nonlinearities, assumption of α  0.1 (between a

Gaussian and a parabolic profile) gives a predicted emittance due to nonlin-

earities alone of εn,max = 0.65 mm mrad, which is roughly half of the full

allowed design emittance given in Table 1.1. An alternative design, which is

discussed below, uses the high gradient (29 MeV/m average) standing wave

(η  1) PWT used at Neptune for acceleration after the gun. In this case, we

have εn,max = 2.75α, which produces a more tolerable margin for emittance

due to nonlinearities and wave-breaking.

2.3 Application to a Real System

In order to see how the concepts discussed in this chapter can be used to un-

derstand the behavior of realistic photoinjector beams with three-dimensional

distributions, an example photoinjector design can be considered. One inter-

esting design example has been developed in the context of the LCLS injector

collaboration [27], in which an ultra-high gradient rf gun (of the same type

used in the Neptune laboratory [34]) is followed by two short (42 cm) PWT

sections to bring the beam to 33 MeV final energy. This design was originally

found by use of a linear slice simulation code, termed homdyn [71]. In the

case of the LCLS photoinjector, the design parameters predicted by homdyn

were verified by multiparticle simulation, as shown below, to be excellent

choices. The design assumes a perfectly uniform beam emitted from the

117
cathode (constant current density up to hard boundaries in radius and time),

and thus is modeled well by homdyn, which assumes the same scenario. On

the other hand, we have found that nonuniformities in both the longitudinal

and transverse charge distributions drive the emittance oscillations (and ir-

reversible growth resulting from transverse phase space wave-breaking) dis-

cussed in this chapter. Thus, in order to illuminate the role of these nonlin-

earities in the space-charge fields, this section compares simulations of this

new LCLS injector design in cases with radially and longitudinally uniform

and nonuniform emission via three-dimensional, self-consistent multiparti-

cle parmela [72] simulations.

The simulation results of the LCLS “working point” design in the case of

uniform transverse and longitudinal beam distributions are shown in Fig-

ures 2.23, 2.24, and 2.25. In this simulation the beam emitted from the

photo-cathode is uniform over 10 psec and a radius of 1 mm. The total charge

is 1 nC, and the peak accelerating fields in the rf gun and PWTs are 140 and

58 MV/m, respectively. In the first of these figures, the transverse (x) rms

beam size and normalized emittance are plotted versus distance down the

beamline. Figure 2.23 illustrates well the two stages of the emittance com-

pensation (that is, the linear inter-slice dynamics discussed in Sections 2.1.1

– 2.1.4) process: the first compensation of the beam from the cathode, us-

ing the solenoid, and the second stage where the beam is matched to the

invariant envelope in both PWTs. Notice that the solenoid connects the two

stages of the compensation process, as it must provide the proper amount of

focusing to produce slices with space-charge dominated waists for the first

compensation point, and it must produce a waist in the rms size of the full

118
 3.0 rms beam size (uniform beam)
rms emittance (uniform beam)

σx [mm], εn,x [mm mrad]

2.0

PWT linacs
1.0

Focusing solenoid
0.0
0 50 100 150 200 250 300 350 400
High gradient RF
photocathode gun
z [cm]

Figure 2.23: The beam envelope and emittance evolution for the LCLS work-
ing point as computed by parmela. The initial beam distribution is uniform
in both the transverse and longitudinal dimensions.

0.02 0.15

0.015
0.1
0.01
0.05
0.005
x [cm]

0 0
βx γ

-0.005
-0.05
-0.01
-0.1
-0.015 (a) (b)

-0.02 -0.15
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.24: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 125 cm after the cathode. The
beam initial distribution is uniform in both the transverse and longitudinal
dimensions.

119
 0.03 0.15

0.02 0.1

0.01 0.05

x [cm]
0 0
βx γ

-0.01 -0.05

-0.02 (a) -0.1

(b)

-0.03 -0.15
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.25: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 400 cm after the cathode. The
initial beam distribution is uniform in both the transverse and longitudinal
dimensions.

beam at the entrance of the first PWT (with σ = σinv ). This coupling of the two

stages further restricts how the system can be operated, and this problem is

studied in Refs. [62, 73].

The first compensation works well in this simulation, as we see that the

emittance begins to decrease immediately after the solenoid. Notice that the

first emittance minimum does not go to zero emittance (or even to the ther-

mal emittance limit, which in this case is roughly 0.3 mm mrad). The reason

for this can be seen in Figure 2.24, which shows the beam’s (x, βx γ) phase

space and (x, z) configuration space at the location of the first emittance min-

imum. Here we see that the leading and trailing beam edges do not follow

the same trajectories as the beam-core slices, as would be expected from the

above analysis for a uniform distribution. This is because the long beam

120
approximation used above is not valid at the longitudinal tail regions, and

the transverse space-charge forces drop dramatically here. As a result, these

regions are strongly focused by the solenoid and are about to go through

cross-over waists.

As just mentioned, a space-charge dominated beam waist must be pro-

duced at the entrance to the first PWT section in order to rms match the

beam onto the invariant envelope. This is because the IE solution specifies

both the beam size and angle, and it has been shown [30] that the impulsive

focusing “kicks” the beam receives in the fringe field regions at the entrance

and exit of the PWT are

γ

∆x
= ∓ x, (2.124)
2γ

where the plus sign corresponds to the defocusing kick the beam receives at

the PWT exit. This kick gives exactly the invariant envelope (x, x
) correlation

provided the beam enters at a waist. Note that the PWT exit kick will cause a

beam matched to the IE to leave the section at a waist as well. This property

— that the IE matched beam enters and exits an accelerator at a space-charge

dominated waist — is a signature of the second compensation, and can be

used to tune simulation parameters. If the next accelerating section is not

too far away, (so that σ and σ
do not change significantly in the drift space

between them) then matching the IE in the first section guarantees matching

it in the second as well. The other signature of the second compensation is

seen in the emittance, which is damped down in the PWTs as γ increases.

Also, as the beam accelerates, the emittance oscillation (plasma) frequency

is quickly decreased.

121
 Figure 2.25 shows the transverse phase and configuration spaces of the

beam at the end of the simulation run. The spatial distribution again shows

that the behavior of the head and tail slices is quite different than that of

the core of the beam. At this point these slices have gone through a cross-

over waist and are expanding. The core slices, on the other hand, have gone

through gentle waists, and are also expanding. Thus, the two bifurcated pop-

ulations have rotated in opposite directions in phase space, and are realigned

near the end of the simulation. At this point the difference between the two

bifurcated populations is that the length in phase space is larger for the slices

in the distribution tails. This effect can also be seen in the Figure 2.23, where

the emittance continues to slowly decrease, even after the second PWT.

The situation will change if nonuniform distributions are used, and we

consider first the case of longitudinal nonuniformity. This case is shown in

Figures 2.26, 2.27, and 2.28, where the beam parameters are the same as for

the uniform case, except the longitudinal profile, which is Gaussian and cut-

off at σz = ±5 psec. As can be seen from the figures, the beam’s behavior in

this case is qualitatively the same as the initially uniform longitudinal profile,

and the evolution of the rms size of the beam is virtually identical to the first

case. The quantitative difference between the two simulations can be seen

in the increased amplitude of the emittance oscillations in the nonuniform

case. This can be seen in Figure 2.26, where the emittance immediately after

the gun and in the first PWT is significantly larger than in the uniform beam

simulation. This increased amplitude is consistent with the linear emittance

compensation theory, as the current of the slices in this case is a function of

longitudinal position, and we expect emittance oscillations to occur.

122
 rms beam size (longitudinal Gaussian)
3.0 rms emittance (longitudinal Gaussian)

σx [mm], εn,x [mm mrad]

2.0

PWT linacs
1.0

Focusing solenoid
0.0
0 50 100 150 200 250 300 350 400
High gradient RF
photocathode gun z [cm]

Figure 2.26: The beam envelope and emittance evolution for the LCLS work-
ing point as computed by parmela. The initial beam distribution is trans-
versely uniform and longitudinally Gaussian, cut off at z = ±σz .

The simulated phase and configuration spaces of the beam at the first

emittance minimum are shown in Figure 2.27. These plots show the effect

of differing slice currents, as the transverse beam extent is a much stronger

function of ζ in configuration space. In phase space, the strongly focused

head and tail slices are about to have ballistic waists, and comparing this

phase space with that of Figure 2.24.a, it is clear that more of the beam parti-

cles follow this cross-over phase space trajectory than in the initially uniform

beam simulation. Like the uniform case, the same phase space realignment

occurs at the end of the simulation, and the phase and configuration spaces

here, shown in Figure 2.28, are similar to those of the uniform beam simu-

123
 0.02 0.15

0.015
0.1
0.01
0.05
0.005
βx γ

x [cm]
0 0

-0.005
-0.05
-0.01
-0.1
-0.015 (a) (b)

-0.02 -0.15
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.27: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 125 cm after the cathode. The initial
beam distribution is transversely uniform and longitudinally Gaussian, cut off
at z = ±σz .

0.04 0.2

0.03 0.15

0.02 0.1

0.01 0.05
x [cm]
βx γ

0 0

-0.01 -0.05

-0.02 -0.1

-0.03 (a) -0.15 (b)

-0.04 -0.2
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.28: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 400 cm after the cathode. The initial
beam distribution is transversely uniform and longitudinally Gaussian, cut off
at z = ±σz .

124
 rms beam size (radial Gaussian)
rms emittance (radial Gaussian)

σx [mm], εn,x [mm mrad]

4.0

2.0

0.0
0 50 100 150 200 250 300 350 400

z [cm]

Figure 2.29: The beam envelope and emittance evolution for the LCLS work-
ing point as computed by parmela. The initial beam distribution is trans-
versely Gaussian, cut off at r = σr , and longitudinally uniform.

lation. Note that the emittance both at the first compensation point and at

the end of the simulation is only slightly higher for the cut-off Gaussian than

for the uniform distribution. This is true because the interslice dynamics are

always reversible.

The dynamical picture changes significantly if one injects a beam with a

radially nonuniform current profile (but the same rms dimensions), as is the

case in the simulation results shown in Figures 2.29, 2.30, and 2.31. Here

a radial Gaussian profile, cut-off at r = σr , is introduced at the cathode.

The rms beam envelope evolution is quite similar to that of the uniform dis-

tribution, as is well known in the theory of space-charge dominated beams

[3]. The phase space dynamics however, reveal many changes from the uni-

form beam case, as can be seen by examining the rms emittance evolution

125
 0.04 0.15

0.03
0.1
0.02
0.05
0.01

x [cm]
βx γ

0 0

-0.01
-0.05
-0.02
-0.1
-0.03 (a) (b)

-0.04 -0.15
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.30: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 125 cm after the cathode. The initial
beam distribution is transversely Gaussian, cut off at r = σr , and longitudi-
nally uniform.

0.08 0.3

0.06
0.2
0.04
0.1
0.02
x [cm]
βx γ

0 0

-0.02
-0.1
-0.04
-0.2
-0.06 (a) (b)

-0.08 -0.3
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

x [cm] z − z [cm]

Figure 2.31: parmela simulations of (a) the transverse phase space and (b)
the configuration space of the beam z = 400 cm after the cathode. The initial
beam distribution is transversely Gaussian, cut off at r = σr , and longitudi-
nally uniform.

126
 0.04 0.04

0.03 0.03

0.02 0.02

0.01 0.01
βx γ

βx γ
0 0

-0.01 -0.01

-0.02 -0.02
(a) (b)
-0.03 -0.03

-0.04 -0.04
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 -0.15 -0.1 -0.05 0 0.05 0.1 0.15

x [cm] x [cm]

Figure 2.32: parmela simulations of transverse phase space of the central

beam slice at z = 125 cm after the cathode in (a) the initially uniform radial
and (b) the radially nonuniform beam distributions.

in Figure 2.29. As expected, the emittance grows significantly when nonlin-

earities in the space-charge field are enhanced by the nonuniformity of the

beam’s radial distribution. The shape of the transverse phase space, shown

in Figure 2.30a, is qualitatively the same as those of the radially uniform

simulations (bow-tie shaped). The reason for this shape is different in the

two cases, however, as the total emittance for the initially nonuniform radial

distribution is dominated by the slice emittance and not by the differential

phase space angles of each slice. This can be seen from Figure 2.32, which

compares the phase space of the central beam slice in the uniform and ra-

dially nonuniform initial distributions at the position of the first emittance

minimum.

The uniform case shows no emittance growth within the slice, while the

nonuniform case does, and it is clear that wave-breaking has occurred, as

127
the tail particles have been focused down into strongly converging trajecto-

ries, and the configuration of the phase space of this slice is similar to that

shown in Figure 2.21. The fact that the slice emittance is much larger in

this case also explains the comparatively shallow first compensation point,

because the alignment of different longitudinal slices has less impact on the

rms emittance of the full beam.

The lack of notably different behavior between the longitudinal beam core

and tails in the nonuniform beam is also seen in the configuration space of

the beam at the end of the simulation, shown in Figure 2.31. Here there

is a clearly defined transverse beam core and halo, which are independent

of longitudinal slice position. This bifurcation is also evident in the phase

space, where the core particles of the beam — the components which have

not undergone wave-breaking — are localized within |x
| < 10 mrad, while

the components which have undergone wave-breaking display large angles

and offsets in x. The core of the beam, which is inside of the initial wave-

breaking point, is well behaved, laminar, and aligned to near the x axis. On

the other hand, the particles which begin outside of the initial wave-breaking

point fold under the core of the distribution, and move quickly across the

x = 0 axis to large amplitude and positive (x, x
) correlation.

The set of simulations of the LCLS photoinjector design given above serve

to illustrate the consequences of the different types of nonuniformities in the

beam distribution, when it is accelerated and transported in accordance with

the methods of the linear emittance compensation theory. In cases where

the nonlinearity of the field is tolerable (as in the perfectly uniform beam

simulations), running the beam essentially on the invariant envelope in a

128
booster linac works well, as predicted by the analysis of SR. In that case, it

can be seen that very few nonuniformities are introduced into the distribution

by the initial emittance oscillation, in which the beam leaves the cathode,

experiences some non-negligible nonlinearities in the field for a short time,

and is then accelerated and focused (to the same radius as at the cathode) to

produce a beam which is appropriate for injection into the linac.

One the other hand, when a moderately nonuniform beam is injected at

the cathode, the initial emittance compensation is degraded and the sec-

ond compensation achieved by matching to the invariant envelope is almost

eliminated — in the uniform beam, the emittance diminishes by 60% in the

second compensation, and in the nonuniform beam case it is diminished by

only 20%. The minimum emittance associated with this process is given by

Equation (2.122), which serves as a useful guide to estimation of the best

performance possible for a given injector configuration.

129
 Chapter 3

Emittance Measurement of

Photoinjector Beams

As argued in Chapter 1, the transverse emittance of an electron beam is an

important factor in parameterizing the beam, as it partially determines the

beam brightness — the other determining factor being the beam current.

Applications of high-brightness beams, such as SASE FELs and advanced ac-

celerators, require not only that the emittance be low, but also that it be well

known — that is, accurately measured. Further, because emittance compen-

sation is such an intricate process, one must in practice tune the system to

achieve low emittance. These facts make the emittance measurement diag-

nostic a critical component of the photoinjector.

The behavior of photoinjector beams, as seen in Chapter 2, can be quite

different than beams produced by other sources. This is because the mecha-

nisms that determine the evolution of these beams are fundamentally differ-

ent from thermal sources, and their behavior is controlled by space-charge

and external forces. The space-charge force couples the beam’s emittance

and current, and as we have seen, the emittance in will change (oscillate) as

it is transported through these systems.

130
 For these reasons, care must be taken in the implementation of the emit-

tance diagnostic used in photoinjectors. In particular, measurement schemes

that deduce a beam’s properties from its behavior in a drift region need to

take into account the space-charge forces that dominantly control the beam’s

behavior. Also, the stability of the system is an important consideration in an

emittance measurement. Fluctuations of the drive laser energy and dimen-

sions may cause shot-to-shot fluctuations in the beam size and emittance

both directly and indirectly (by changing both the emittance compensation

process, and the amount of space-charge present in the region of the mea-

surement.) For this reason, single shot measurements are preferable to those

requiring many shots.

In this chapter, two emittance measurement schemes are reviewed and

examined in terms of the effect of space-charge — the multislit-based and

quadrupole (quad) scan systems. The subject of emittance measurement

systems has been studies previously in general circumstances [5] and for

photoinjectors in particular [74]. Here, some of the relevant issues in Refs.

[5] and [74] are discussed and the key differences between this and previ-

ous work are clarified. These two emittance measurement techniques are

compared in an experiment performed at the Lawrence Livermore National

Laboratory (LLNL) Thompson source photoinjector [75], which offered a 5

MeV, highly space-charge dominated electron beam suitable for this study.

The results of this experiment, which show that the emittance found with the

quadrupole scan is in disagreement with that measured with the multi-slit

system (the quad scan results are consistently higher), are then compared

with simulation and analytical models.

131
 Space-charge Emittance Dominated Beamlet Intensity
Dominated Beam Slits Beamlets Profile

xm σm

Figure 3.1: Illustration of the multislit-based emittance measurement

scheme.

3.1 Emittance Measurement Techniques

In this section, the two emittance measurement techniques used in the LLNL

experiment, the slit based system and the quad scan, are introduced. In

addition to introducing the operating principles of these methods, the role

space-charge plays in the measurement schemes will be discussed, in terms

of how it may affect the emittance value derived from the measurement.

3.1.1 Multislit-based Measurement

The first purpose of the multislit collimation system, which is actually a trace

space measurement system, is to slice up the beam into physically separated

line-like sources, or beamlets, as shown in Figure 3.1. This slicing combined

with a drift, which reveals the spread in velocities as spatial information

132
at an intensity-sensitive detector, allows a full reconstruction of one of the

beam’s transverse trace plane distributions, e.g., (x, x
). This technique is

especially useful when implemented as a single-shot diagnostic in photoin-

jectors, where shot-to-shot beam parameter fluctuations may dominate mul-

tishot measurements. Multishot variants of the slit-based method involving

one or two slits, as well as the pepper-pot method (the two-dimensional ana-

log on the multislit system, which can give the four-dimensional transverse

trace space), are discussed in detail in Ref. [5]. The method of determining

the trace space from the multislit technique is introduced here, and the use

of the multislit array to mitigate collective effects in space-charge dominated

beams (which can be large enough to invalidate measurements) is examined

in depth, as these effects are our primary concern.

The beamlets are created by the multislit mask at a constant spacing w,

which is much larger than the slit width d. The beamlet distribution is then

detected downstream of the multislit mask, and the beamlets resolved. The

width of each beamlet gives a measure of the width of the transverse mo-

mentum distribution at each slit, and the centroid of the beamlets gives the

correlated offset of the momentum distribution of each slit. Assuming a drift

length L between the multislit mask and the detecting plane, the measured

trace space distribution is approximated by a series (m = 1, . . . , N) of dis-

tinct angular distributions at regularly spaced (take this dimension to be x)

intervals. The beamlets yield the correlated beam divergences

xm − xm,c  xm − mw

xm,c = = , (3.1)
L L

133
and the rms spreads in divergence



2 /L2 − x

σm = xm 2
m,c , (3.2)

where the average  is performed over the distribution in the mth beamlet.

Here it is assumed that the final spread in detected beamlet size is much

greater than the slit width. Once these parameters are extracted from the

data, a graphical trace space distribution can be constructed, as illustrated

in Figure 3.2.

Note that the trace space distribution is centered in both x and x
, by

subtracting off the overall value of the centroids,

)N

 m=1 Im mw
xm,c = xm,c − xc old = mw − )N , (3.3)
m=1 Im

and
)N
   

 m=1 Im xm,c


xm,c =

xm,c − xc
old = xm,c old −


)N , (3.4)
m=1 Im

where Im is the integrated detected intensity of the mth beamlet.

The second moments of the trace space distribution are obtained from

the parameters given by Equations (3.3) and (3.4) as

)N 2
m=1 Im xm,c
x  =2
)N , (3.5)
m=1 Im

)N  


2
m=1 I m xm,c
2
+ σ m
x
2  = )N , (3.6)
m=1 Im

and
)N


m=1 Im xm,c xm,c
xx  = )N , (3.7)
m=1 Im

where, in obtaining the moment x
2 , it is assumed that the beamlet distri-

butions are symmetric about their centroids. From the second moments, the

134
 2
1.5
1

x
[mrad] 0.5
0
-0.5
-1
-1.5 (a)
-2
-3 -2 -1 0 1 2 3
x [mm]

2
x
[mrad]

(b)
-2
-4 0 4
x [mm]

Figure 3.2: (a) Beam trace space constructed from the beamlet intensity pro-
file illustrated in Figure 3.1. Each point represents the position of a beamlet in
trace space and the error bars indicated the thermal spread of the beamlets.
(b) Contour plot representation of the same data. Here the relative weights
of the beamlets are distinguished by the color in the plot, with red indicating
the highest intensity, violet the lowest.

135
 
rms emittance is found using the standard definition, εx ≡ x 2 x
2  − xx
2 ,

as well as the other rms Twiss parameters [76],

x 2 
βx ≡ ,
εx
xx

αx ≡ − , (3.8)
εx
x
2  1 + α2x
γx ≡ = .
εx βx

The slits serve a secondary purpose, which is of central importance in

the discussion of photoinjector beams, since these beams are space-charge

dominated for almost all relevant energies and beam sizes. The notion of

space-charge dominated flow is quantified, as indicated previously, by com-

paring the space-charge and emittance terms in the rms envelope equation in

a drift space (that is, by setting the acceleration and external focusing terms

to zero in Equation (2.53))

2
εn I
σx

= +
, (3.9)
γ 2 σx3
γ I 0 σx + σy
3

where the normalized emittance, εn  γεx , is used, and, of course, an analo-

gous equation can be written form σy . Taking the ratio of the second to the

first terms on the right-hand side of the envelope equation, and assuming a

round beam (σx = σy = σ0 ), we have a measure of the degree of space-charge

dominance over emittance in driving the evolution of the beam envelope,

Iσ02
R0 = 2
= 2k2p β2x . (3.10)
2I0 γεn

In order to illustrate the relationship between the two effects driving the beam

envelope, this ratio is written in terms of the beam plasma wave number kp ,

and the rms beta function βx .

136
 As an example, consider parameters typical of the experiment described

below: a 5 MeV electron beam with current of 100 A, rms beam size of 1

mm, and normalized rms emittance of 4 mm mrad. This yields a ratio of

R0  75, and one can see that this beam cannot be emittance dominated

until it is focused down to small sizes, σ0 < 100µm. Thus linear transport

theory cannot be used to measure the emittance with this type of beam, as will

be discussed in evaluating the quadrupole scanning technique. Collimation

with slits mitigates this situation, however, by creating low current, small σx

beamlets that have the same uncorrelated transverse momentum spread as

the original beam. Noting that the rms size of a uniform beamlet created
√
by a vertical slit of width d is σx = d/ 12, and assuming σx  σy , the

space-charge dominance ratio of the beamlets is

  2
2 I d
Rb = . (3.11)
3π γI0 εn

For the beam parameters of this example, and the choice in the exper-

iments of d = 50µm, Rb = 0.042  1, implying that the beamlets which

pass out of the slits are emittance dominated. This process can also be

understood in terms of the plasma frequency of the beam and the beta

function. Since the beamlets have the same density as the beam before the

slits, kp is unchanged by collimation. On the other hand, the beta function

of each beamlet is much smaller than that of the beam before collimation,
 √ 
βx  σx /σx
→ d/ 12 /σx
, and for the beamlets the ratio Rb = 2k2p β2x,b

can therefore be made much less than unity.

There are many further technical and physical consideration which must

be taken into account in order to optimally design a slit-based trace space

137
measurement system [56]. These considerations determine the geometry of

the slit mask and optimal drift length between the slits and profile moni-

tor, and thus determine the resolution of the measurement system. Since

the main concern of this chapter is the overall effect of space-charge on the

results of emittance measurements, these ancillary concerns are not inves-

tigate further here. Ref [77] discusses issues specific to the mulislit system

used in this experiment, and more general material on the optimization of

this type of instrument is found in the comprehensive review of emittance

measurement techniques given in Ref [5].

3.1.2 Quadrupole Scanning Emittance Measurements

The quadrupole scanning technique for measuring emittance is well known

and widely used in the accelerator physics community [78]. A brief descrip-

tion of the process is given here in order to show its limitations when applied

to space-charge dominated beams. One can understand the measurement by

considering the evolution of the beam’s rms Twiss parameters by differen-

tiating Equations (3.8) in a drift length after a thin focusing lens (of focal

length f )

xx

β
x = 2 = −2αx ,
εx
x
2  + xx


α
x = − = −γx , (3.12)
εx
x
x


γx
= 2 = 0,
εx

where x

= 0 is used when there are no external forces, and the space-

charge force is ignored. Using Equations (3.12) and applying the thin lens

138
(αx0 becomes αx0 + βx0 /f ) gives

βx0
βx (z) = βx0 − 2 αx0 + (z − z0 )
f
   
βx0 2 (3.13)
 1 + α x0 + f  2
+  (z − z0 ) .
βx0

Rearranging Equation (3.13) in terms of 1/f and multiplying by the emit-

tance gives an equation for the square of the beam size as a function of the

focusing strength of the lens,

% &
σx2 (z) = σx0
2
− 2αx0 εx (z − z0 ) + γx0 (z − z0 )2
2
 
2σx0 αx0 2
+ (z − z0 ) − (z − z0 ) (3.14)
f βx0
2
σx0 2
+ (z − z0 ) .
f2
2
This equation is quadratic in 1/f (a (1/f ) + b (1/f ) + c), and using these

coefficients, the emittance is given by

b2
εx2 (z − z0 )4 = ac − . (3.15)
4

With this analysis in mind, the emittance can be obtained by measuring the

beam size at a given drift length Ld after a quadrupole magnet, scanning

through a range of focusing strengths. The same result could have been

derived by solving the envelope equation for a drifting beam without space

charge
εx2
σx

= . (3.16)
σx3

It is important to stress here that the quad scan formalism is based on

rms quantities. That is, Equation (3.14) holds (without space-charge) for ar-

bitrary, evolving beam distributions, provided that we are measuring the rms

139
value of the beam size. When considering the effect of space charge in the

remainder of this chapter, we are looking for gross differences in the rms

beam size. This point is seen also in the envelope equations including space-

charge. In these equations, as was shown in Chapter 2, the space-charge term

applies for any form of the beam’s distribution function because the rms size

is affected by only the linear part of the space-charge force (defined in the

least-squares sense). This study expands on that of Ref [74], which points

out that the assumption of a Gaussian transverse profile and measurement

of full width at half maximum (FWHM) spot sizes is incorrect for nonther-

malized, photoinjector beams. For these reasons, all experimental spot sizes

quoted here are determined using rms measurements of the distribution,

thus avoiding potential problems Gaussian fitting and FWHM-based analyses

of the data.

We can now examine how a quadrupole scan behaves in the other ex-

treme, namely the one where we assume the beam has space-charge but no

emittance. In this case, the envelope equations in x and y become

I
σx

=
,
0 σx + σy
γ 3I
(3.17)
I
σy

= 3
.
γ I 0 σy + σx

These equations have the same form as the the radial equations considered

in Chapter 2, and thus have no exact solutions. The usual approximation,

applied to this situation gives σx + σy ≈ constant. This approximation is

acceptable as long as kp Ld < 1. Using this and a change of variables (Σ =

140
σx + σy , and ∆ = σx − σy ) gives

2I
Σ

= ≈ 2k2p Σ0 ≈ constant,
γ3I 0Σ
(3.18)



∆ ≈ 0.

This set of equations is easily solvable, and the solution is

Σ (z) = Σ0 + Σ
0 z + k2p Σ0 z2

(3.19)
∆ (z) = ∆0 .

1
Using σx = 2 (Σ + ∆), and assuming an initially axisymmetric beam (σx0 =

σy0 ), allows a solution of the x envelope equation,

σx0
σx (z) = σx0 + σx0 − z + k2p z2 σx0 . (3.20)
f

2
The result that the beam size is proportional to kp z is no surprise, as it

has been seen several times when considering space-charge dominated flow

in Chapter 2. It is interesting however, that the rms beam size is linear with

the focusing strength of the lens, and therefore, σx2 is quadratic in 1/f , just

as in the case where space-charge is ignored. If one attempts to compute

the emittance using the coefficients of powers of 1/f , as in the emittance

dominated case, the result is zero. This is understandable in the sense that

one would expect the beam size to remain linear with 1/f all the way down

to zero thickness for a beam with no emittance.

For the measurements described below, neither of these limiting cases is

applicable as both the emittance and space-charge terms are important in

the evolution of the rms beam size. What is shown here, however, is that one

can expect the qualitative outcome of the quadrupole scan to be the same in

either regime. In particular, it is notable that, even in the case of space-charge

141
dominated beam dynamics, the algorithm used to extract the emittance from

a quadrupole scan [based on Equation (3.15)] gives a well-behaved result, and

there is no a priori reason that data from these scans would be rejected as

unphysical. The experimental data shown below demonstrate quad scans

that indeed yield good fits to Equation (3.15), but which have systematic

errors in the resulting calculated emittance. The basis for these errors, which

is dependent on the interplay between space-charge and emittance forcing

of the beam envelope, is discussed below, in the context of the experimental

data and modeling.

3.2 Experimental Setup and Procedure

The quad scan and multislit-based emittance measurement techniques were

compared using the Thomson scattering photoinjector at Lawrence Liver-

more National Laboratory. The beam line configuration used for these mea-

surements is shown in Figure 3.3. The accelerator in this setup was a 1.6 cell,

S-band, BNL-SLAC-UCLA-LLNL rf photo-cathode gun [21], of the same basic

design as the gun employed at Neptune. This gun produced a 5 MeV beam

whose charge, transverse, and longitudinal spot sizes varied as given in Ta-

ble 3.1, and described below. An emittance compensating solenoid after the

gun was used to control the beam size after the gun, which allowed the se-

lection of a reasonable, and well diagnosed beam size at the emittance slits

and the quadrupole. The magnetic field at the cathode was nulled with an

identical bucking solenoid placed upstream of the cathode. The charge was

measured using an integrating current transformer, which, in the case of the

142
 Quadrupole

YAG Screen

Slits

Figure 3.3: The LLNL Thomson source photoinjector beamline used to mea-
sure the emittance with the quad scan and slit-based techniques.

Parameter Range

electron beam charge 50–300 pC

rms laser spot size 0.5–2 mm

rms laser pulse length 2.5–6 psec

Table 3.1: Electron beam parameter range used for emittance measure-
ments.

slit measurements, allowed the measurement of the charge and emittance

in a single shot. The laser injection phase was monitored during the exper-

iments by mixing the low level rf derived from the laser oscillator with that

from a probe in the gun full cell. Long time scale drifts in the rf phase were

corrected for all measurements with a manual phase shifter.

From Equation (3.20) we expect the plasma wave number to be a key pa-

rameter in the quad scan measurement of the space-charge dominated beam.

In order to make this parameter vary in the experimental runs, the dimen-

143
sions of the photo-cathode drive laser were made to vary, thus changing the

beam density and the plasma frequency of the beam. To indicate how the

laser dimensions were adjusted, a description of the drive laser system is

given here.

A Ti:sapphire-based laser system was used to produce the UV pulses nec-

essary for photoelectron emission in the gun. The photoinjector laser system

(PLS) is seeded by stretched pulses from the Falcon laser [79], so that the pho-

toelectrons and the rf system of the linac can be synchronized to the Falcon

master oscillator. The seed light is introduced into a regenerative ampli-

fier (regen) of the photoinjector system through a single-mode, polarization-

preserving fiber that runs from the Falcon stretcher output to the laboratory

containing the PLS. In addition to the light pulse, timing signals are sent from

the Falcon laser system to PLS to match the drive laser timing with the Fal-

con laser and linac rf system. The PLS consists of a fiber-seeded regenerative

amplifier, a multipass power amplifier, a pulse compressor, and frequency

conversion crystals for frequency tripling the 800 nm laser pulse. The two

laser amplifiers, the regen and multipass, are pumped by a single, frequency-

doubled, Q switched YAG laser that puts out 300 mJ of 532 nm light in an

8 nsec pulse. The output beam of the multipass amplifier is expanded and

sent to the grating compressor.

The pulse width at the output of the grating compressor was measured us-

ing a single-shot autocorrelator. The desired pulse length of the UV pulse was

a few psec, which can be obtained by varying the distance between the grat-

ings in the compressor. We attempted to directly measure pulse width versus

grating separation but found that the autocorrelator gives good results only

144
for pulses of length ≈ 1 psec or less. Consequently, pulse lengths greater

than 1 psec were estimated using various techniques. For pulse lengths less

than 1 psec, the autocorrelator appeared to give accurate results, and the

shortest pulse measured was 150 fsec FWHM, which corresponds to a time-

bandwidth-limited pulse.

By varying the difference in grating separation from the minimum pulse

length position ∆b between the gratings in the pulse compressor the pulse

length was continuously adjustable (in the IR) from 180 psec to 150 fsec. This

variation was used to change the beam intensity as it was launched at the

photo-cathode, thus also varying the plasma frequency of the beam. Three

separate methods of modeling the compressor all gave similar results, which

were that the grating separation tuning, ∆T /∆b, was found to be about 0.2

psec/mm in the IR (or at 1ω). The UV (3ω) intensity is proportional to the

1ω intensity cubed, and, as a result, a longitudinally Gaussian pulse narrows

in time, i.e.,
∆T1ω
∆T3ω = √ . (3.21)
3
Based on this, the expected pulse width of the 3ω light was given by

∆T3ω = (0.11psec/mm) × ∆b (mm) . (3.22)

For the multislit-based emittance measurements, the slit array was cho-

sen to be a set of stainless steel collimating slits with a 50 µm slit width, 0.75

mm separation, and 5 mm depth. The drift length from the slits to the mea-

surement screen was approximately 50 cm. The choice of these parameters

was based on the expected electron beam parameters, the considerations

given above, and the design considerations of a slit-based system given in

145
Ref. [52]. For a beam size of 1-2 mm on the slits and at 5 MeV, this allowed

us to measure a maximum normalized emittance of 12 mm mrad using the

slits. The slits were mounted on an insertable, rotatable actuator, powered

by stepper motors. This allowed the alignment of the slits to the beam, insur-

ing the proper angle for maximum acceptance of the slits. A 0.5 mm thick

YAG:Ce crystal was used as the intercepting screen for both the slit-based

and quad scan measurements. The beam images produced by the crystal

were captured by a video camera and digitized by a computer controlled

frame grabber. Once captured, the beam images were analyzed on-line in the

case of the multislit measurements, and saved for latter analysis in the case

of the quadrupole scan.

Given the apparatus described above, the experimental procedure was

to set the compressor gratings to a given separation distance and measure

the dimensions of the laser and electron beams as well as the charge and

injection phase. This was important because it allowed the calculation of the

plasma frequency of the beam in order to compare our measurements with

simulation and analytical models. Once these parameters were known, the

beam emittance was measured using both the multislit and quad scanning

techniques. Table 3.1 lists the range in beam parameters over which the

emittance measurements were done.

3.3 Experimental Results

The quad scan and multislit measurements were performed for seven dif-

ferent electron beam pulse lengths. Figure 3.4 shows one image of a beam

146
horizontally focused by the quadrupole. The horizontal rms size of the beam

in this image is 140 µm. The figure also shows a representative result of one

of the quadrupole scans. Figure 3.5 shows the data found using the slits

for the same grating separation as used for the quad scan of Figure 3.4. The

emittance calculated via the quad scan measurement is higher than with the

emittance slits. This pattern was repeated in all of the measurements we

performed.

As argued above, the result of the quadrupole scan is expected to depend

not only on the emittance but also the strength of the space-charge defocus-

ing forces encountered by the beam through the drift region. For this reason

the results of the measurements are plotted in Figure 3.6 as a function of the

plasma wave number of the beam at the quadrupole multiplied by the drift

length between the quadrupole and the screen. The plasma wave number is

determined through simulation, using the measured properties of the laser

and the total charge launched, as discussed further in the following section.

For this range of beam parameters, the quad scan gives consistently higher

values for the emittance than the multislit measurements.

3.4 Data Analysis

The recording of laser spot size, length, and energy for each set of emit-

tance measurements allowed the accurate simulation of the acceleration and

focusing of the beam through the 1.625 cell gun. These simulations were

done using parmela. In these simulations, a beam derived from injection

of a Gaussian longitudinal and transverse laser pulse shape was used. The

147
 1
y = ax 2 + bx + c
0.8
a = 8.3828
0.6 b = −6.4227

σx2 [mm2 ]
c = 1.238
0.4

0.2 (b)
(a)
8.46 mm 0
1.5 2 2.5 3 3.5

1/f [1/m]

Figure 3.4: (a) False color image of a beam used in a quadrupole scan. (b)
Result of one quad scan. The normalized horizontal emittance found from
the curve fit is 9.6 ± 1.1 mm mrad.

45

40
Relative Intensity

35

30

25

20
(a)
15 (b)
11.0 mm
10
0 2 4 6 8 10

x [mm]

Figure 3.5: (a) False color image of a beam passing through the collimating
slits. (b) Intensity graph found by summing the vertical pixel values at a
given horizontal position. The normalized horizontal emittance calculated
from this plot is 6.9 ± 0.7 mm mrad.

148
 20

Slit Emittance
Quad Scan Emittance
15

εx,n [mm mrad] 10

0
0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
kp Ld

Figure 3.6: Plot comparing the different emittance measurement techniques

for varying beam intensities. The strength of the space-charge forces is
parameterized in the scan by the product of the drift length between the
quadrupole and the detector and the plasma wave number at the quad.

purpose of these simulations was both to get a base line expectation value

for the emittance of the beam and to find the spatial dimensions of the beam

at the position of the quadrupole. The results of the parmela simulations

are shown in Figure 3.7. The figure shows that the emittance predicted by

parmela is generally slightly lower than that of the multislit measurements,

but follows the same trend from measurement to measurement. This indi-

cates that the small variation in the multislit emittance values at different

beam densities is due, at least in part, to changes in the emittance compen-

sation process caused by different beam densities at the cathode, as could

be expected.

149
 20
parmela
Quad Scan Emittance
15 Slit Emittance

εx,n [mm mrad] 10

0
0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
kp Ld

Figure 3.7: parmela simulations of the emittance at the quadrupole and the
measured emittance using both measurement methods.

At this point the quadrupole scan procedure was simulated with space-

charge forces included in the computations. This was done in three different

ways. The first was to use the parmela beam distributions from the previ-

ous simulations at the position of the quadrupole and iteratively simulate

their propagation in the drift region between the quad and the measurement

screen for different quad strengths. In these parmela runs, we changed the

space-charge routine to a point-to-point Coulomb method, which is generally

less accurate than the standard radial algorithm, but was needed to model

the beam when it becomes highly asymmetric after quad focusing. In order

to avoid the noise inherent in the point-to-point calculations, we also em-

ployed the slice code homdyn. Finally, we also simulated the quad scans

by numerically integrating the two dimensional envelope equations for the

150
 20 Slit Emittance
Simulated Quad Scan Emittance
Quad Scan Emittance
parmela
15
εx,n [mm mrad]
10

0
0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
kp Ld

Figure 3.8: homdyn simulation of the quadrupole scan plotted with the
parmela simulations and the measured emittance data.

beam including space-charge,

2
εn I
σx

= +
,
γ 2 σx3
γ I 0 σx + σy
3
(3.23)
ε2 I
σy

= 2n 3 + 3
.
γ σy γ I 0 σx + σy

In the case of both the homdyn and envelope equation simulations, the

emittances εn [cf. Equation (3.23)] were taken to be that measured by the

slits, the initial beam sizes were taken from parmela, and the beam cur-

rents taken from measurement data. In all three cases the simulations gave

similar results, and the less noisy (but still three dimensional, and sensi-

tive to differential slice dynamics [62, 80]) homdyn results are shown in

Figure 3.8. In order to avoid confusion about the meaning of the emittance

term in Equation (3.23), we specify at this point that the symbol εn will in-

151
dicate the normalized thermal emittance. The thermal component to a full

measured emittance is due to uncorrelated motion, as opposed to the pro-

jected emittance observed from residual correlations between longitudinal

and transverse phase space in poorly compensated beams. Therefore, in the

integration of Equation (3.23) it is assumed that the measured multislit emit-

tance is roughly equal to the slice emittance. This assumption is justified for

two reasons. First, because we launch a transversely Gaussian distribution

from the cathode, there is a large amount of nonlinear space-charge field

induced slice emittance growth, as seen in the previous chapter. Second,

the location of the slits roughly corresponds to the emittance compensation

point, and thus the rms emittance of the beam is dominated by the slice emit-

tance at this point. The assumption that the rms emittance is dominated by

that of the slices is also seen in the slit emittance data, where different beam

intensities give only minor differences in the measured emittance. This result

would not occur if the slice and projected emittances were vastly different,

since changing the beam density would change the position of the emittance

compensation point and thus the emittance measured by the slits. This dis-

tinction is physically meaningful, as the interplay between space-charge and

emittance described below gives rise to the observed effects precisely because

the emittance is thermal.

As Figure 3.8 shows, the emittances derived from the simulated quad

scans are indeed higher than the actual emittances used in the simulations,

as expected. In addition, they agree reasonably well with the results of the

quad scan measurements.

The parameter kp Ld was chosen to plot against the emittance results be-

152
cause it describes the degree to which the plasma nature of the beam in-

fluences its motion in the experiments. In cases where the beam evolves in

a purely space-charge dominated way (such as the emittance compensation

process), kp Ld /2π would be the number of plasma oscillations that the beam

undergoes. However, in the case of the quad scan, the beam’s size at the end

of the drift is also affected by emittance, especially at points where the beam

has gone through a ballistic waist. In that case, the relative strength of the

emittance and space-charge terms (R0 ) in the envelope equation are of in-

terest. The ratio of these forcing terms can be characterized by the beam’s

beta function times its plasma wave number (in fact, Equation (3.10) says

that kp βx = R0 /2). Thus, there two parameters that may effect the result

of the quad scan: kp Ld , which indicates the amount of time over which the

space-charge forces are allowed to act, and kp βx , which determines the rela-

tive strength of the space-charge to emittance terms. In Figure 3.9 the path of

the experimental data in the plane defined by kp βx and kp Ld is plotted. The

background of the plane is a contour plot of simulated quad scan emittances

using the envelope integration model, with a constant thermal emittance for

each point in the plane. The figure shows that some data points that hap-

pen to be close together in kp Ld have significantly different values of kp βx .

This effect explains in part the drastic difference in emittance between close

data points in Figure 3.8. Note also that the lower right-hand corner of the

graph has values of emittance that are lower than the value that was used to

integrate the envelope equations. This point will be discussed further below.

It is instructive at this point to look again at the quad scan procedure to

see why the data shows, and simulations predict, erroneous numbers for the

153
 20 12

10

15

εx,n [mm mrad]

kp βx

10

5 4
0.6 0.8 1

kp Ld

Figure 3.9: Contour plot of the simulated quad scan emittance over a range
of both kp Ld and kp βx . The white plot points locate the positions of the ex-
perimental data. The normalized emittance used as input to the simulations
was 5 mm mrad, while the output emittance values rage from 4 (deep blue)
to 12 mm mrad (red).

emittance. In particular, we can see both in the data and in simulations that

there is an asymmetry in the curves about their minima. In Figure 3.10 the

quad scan data from Figure 3.4 is replotted, and the data points on either

side of the minimum are fitted to different curves. This figure shows that

the points before the minimum spot size follow a path with less curvature

than the points occurring after the focus. The second part of the figure

shows this effect more dramatically. In that case the plot points come from

a simulation of a quad scan assuming conditions similar to the experiment,

154
 1 1.2

0.8 1

0.8
0.6

σx2 [mm2 ]
σx2 [mm2 ]

0.6
0.4
0.4

0.2
0.2
(a) (b)
0 0
1.5 2 2.5 3 3.5 1 2 3 4 5 6 7

1/f [1/m] 1/f [1/m]

Figure 3.10: (a) Quad scan data as shown in Figure 3.4b. Here different fits
are applied to the data before and after the focus. (b) Simulation of a quad
scan with an extremely space-charge dominated beam. Again, the two fits on
either side of the minimum spot size are shown to illustrate the asymmetry
in the simulation points.

but with a higher beam current of several hundred amps.

This asymmetry in the data about the minimum spot size is a manifes-

tation of the fact that the evolution of the beam through the drift is very

different for data points on opposite sides of the minimum. For points on

the right-hand side of the curve (weaker focusing) the beam size is deflected

appreciably only by space-charge. For points in the curve at and to the left

of the beam waist, corresponding to stronger quad focusing, there has oc-

curred at some position a thermal emittance dominated beam waist. In the

region of that waist, the emittance force “turns on” and applies an extra kick

to the beam size that deflects it away from the path it would take due only to

space-charge. The stronger the quadrupole focuses (larger 1/f ), the stronger

this thermal emittance kick will be, and the further upstream of the measure-

ment screen it occurs. Figure 3.11 illustrates this behavior, by showing the

155
 2.5
Emittance included
2 Space-charge only

1.5

0.5

0
2 3 4 5 6 7

Figure 3.11: Simulations of a quad scan with and without emittance. The
solid line represents a simulation without the emittance forces, while the
data points show a simulation of the full envelope equations.

results of a simulation with and without the emittance term included in the

envelope equations. Notice here that for points on the left-hand side of the

minimum, the simulations with and without emittance agree very well, indi-

cating that the integrated motion of the beam is indeed purely dominated by

space-charge. However, the points at and after the minimum diverge rapidly

from the path that the simulation with space-charge along predicts, even

though the beam is space-charge dominated over most (but not all) of its

trajectory.

It is interesting to note that both a purely thermal emittance dominated

and a purely space-charge dominated scan would produce a (σx2 , 1/f ) curve

which is symmetric about the minimum σx . The asymmetry observed in the

mostly space-charge dominated quadrupole scan is a result of both space-

156
charge and thermal emittance effects asserting themselves in the measure-

ment.

Given this problem, in which points on one side of the curve are space-

charge dominated, while on the other side they are affected by emittance, it

is certainly no surprise that the emittance computed from these curves is not

correct. In addition, the asymmetry of the curve introduces a problem in the

consistency of the result. That is, since the computation of the emittance

requires us to fit a parabola to a curve that is of higher order, the fit parame-

ters will depend on the portion of the curve used for the fit. To illustrate this

point, consider a single quad scan simulation as shown in Figure 3.12a. The

range of focal lengths we use for the fit is in principle arbitrary; however, it

is reasonable to impose the conditions that the simulation points are equally

spaced in 1/f , and that the end points give the same beam size, which is

chosen here to be about one mm. These are the conditions used for simu-

lations in this study. One can see how the computed emittance varies when

the starting and stopping points of the fit are varied. Figure 3.12b shows

the emittances calculated from the simulation data of Figure 3.12a with the

horizontal and vertical axes representing the starting and stopping points of

the fit, respectively.

There are two noteworthy features of Figure 3.12b. The first is that it

shows both regions above and below the input value of the thermal emittance.

In fact, the violet regions in the upper half of the plot represent points where

the computed emittance is imaginary (this happens if the right-hand side as

determined from the scan parameters of Equation (3.15) is negative). The

other point is that the upper left corner of the graph is the region where the

157
 3.5

2.5
σx2 [mm2 ]
2

1.5

0.5 (a)

0
0 1 2 3 4 5 6

1/f [1/m]

10
5
1/f [1/m] (ending point)

εx,n [mm mrad]

4
6

3
4

2 2
(b)
1 0
1 2 3 4 5
1/f [1/m] (starting point)

Figure 3.12: (a) Quad scan simulation used to perform the fitting in (b). The
thermal emittance used in the simulation is 4 mm mrad. (b) Contour plot of
the emittance calculated by fitting the data in (a). The portion of the graph
below the diagonal is unused (this half would be the mirror image of the
upper half). Here, red represents an emittance of 10 mm mrad and violet is
imaginary (the square of the derived emittance is negative).

158
gradient in emittance values is the steepest. This is also the region that was

initially assumed to be the most reasonable. The lower left-hand corner of

the plot seems to be a stable region, but it is not an accurate solution. Points

in this corner fit only the first part of the curve, as shown in Figure 3.11. This

strict dependence of the computed emittance on the range of fit points is

another factor that may help to explain the inconsistency in the quad scan

measurements. This effect was not considered at the time the measurements

were performed, and for the majority of the quad scans, slightly more data

points were taken before the minimum of the curve than after.

3.5 Conclusions

Two different emittance measurement techniques were discussed in this chap-

ter: the multislit-based measurement and the quadrupole scan. For the

highly space-charge dominated beams of interest here, the slit collimation

of the beam into beamlets reduces the quantity kp βx sufficiently to allow the

expansion of the beamlets to be thermal emittance dominated. In the experi-

ment performed at the Thomson source photoinjector, it was found that the

emittances measured with the multislit system were reasonably independent

of the beam intensity, and agreed well with the parmela simulations.

In contrast, the quadrupole scanning procedure was found to be ill-suited

to measuring the emittance of these highly space-charge dominated beams.

In addition to the issues associated with multishot measurements in systems

with significant shot-to-shot charge fluctuations, the fundamental problem

with the quadrupole scan is that the beam evolves under the influence of

159
both space-charge and emittance effects. In the Livermore experiments it

was found that the emittance measured with the quad scan was consistently

higher than the multislit measurements and parmela predictions. In addi-

tion, simulations of the quad scan for the beam parameters of the measure-

ments also show higher values for the emittance, and reproduce the system-

atic dependence of the measured emittance on the quantity kp Ld .

Simulations show that when kp Ld is of the order of unity or greater, space-

charge forces are large enough to significantly alter the evolution of the beam

in the drift region of the scan. Further, when the quantity kp βx is larger than

unity (indicating the degree to which the space-charge forces are dominant

over the emittance effects), the quadrupole measurement of the emittance

will have have notable errors. For kp βx  1, the emittance measurement is

no longer valid at all, but is really only a measure of the intricate interplay be-

tween the space-charge and emittance effects during the focused trajectory.

Thus the conditions

kp Ld > 1 and
(3.24)
kp βx > 1

are theoretical tools for determining the range of parameters over which the

validity of the quadrupole scan is no longer guaranteed. Perhaps an even

more useful method that can test for possible problems in a quadrupole

scan is the examination of the scan data itself. When the (σx2 , 1/f ) curve is

no longer symmetric in 1/f about the minimum in σx2 , this is a clear signa-

ture that the combined space-charge/emittance dominated beam evolution

produces an unreliable measurement. It is both inaccurate and imprecise, as

the exact value of the emittance derived from the (σx2 , 1/f ) curve depends on

160
the number of points on either side of the minimum that are used in the fit.

It may be noted that, as experimental progress is made in lowering the

beam emittance obtained from rf photoinjectors, the problems described

here with quadrupole scans will be exacerbated. In recognition of such dif-

ficulties, most laboratories use slit-based measurements for low energy (<10

MeV) emittance diagnosis in rf photoinjectors. The present results imply

[using Equations (3.24)] that quad scans may be problematic at energies sig-

nificantly higher than 10 MeV.

161
 Chapter 4

Magnetic Compression

and Emittance Growth

Up to this point, this dissertation has concentrated on one aspect of the topic

of high brightness beams, that is the physics of low emittance beams in pho-

toinjectors. The other aspect of this topic is the creation of short pulse —

high current — beams. This chapter examines one method of producing a

very short, high current photoinjector beam using a magnetic compressor.

The motivation for creating this type of beam is again supplied by advanced

high brightness beam applications, which require rms pulse lengths shorter

than one pico-second [8]. The rf photoinjector is a natural choice of source for

this kind of beam because, as discussed in Chapter 1, its intrinsic brightness

is extremely high. On the other hand, this high brightness has consequences

that put a lower limit on the length of the beam produced. The biggest con-

sequence of the intensity of the beam emitted from the photo-cathode is its

plasma-like behavior, driven by the beam’s space-charge forces. As shown in

Chapter 2, the way to control the beam’s space-charge forces, and therefore,

preserve its initial brightness, is through the emittance compensation pro-

cess. This process determines the necessary plasma frequency, and thus the

162
density, of the beam by geometric and external field considerations. In prac-

tice, emittance compensation sets the acceptable length of a photoinjector

beam to be in the range of several pico-seconds or more, and it is not uncom-

mon at such current levels for some space-charge induced pulse lengthening

to occur [81]. In order to obtain the currents (and pulse lengths) needed by

applications, therefore, longitudinal focusing, or compression of the beam

after the injector is required [39].

The most commonly employed method of pulse compression uses a mag-

netic chicane [82], such as the Neptune chicane described in Chapter 1. This

magnet configuration compresses the beam when used in conjunction with

off-crest acceleration in an rf linac. Running the electron pulse off-crest —

ahead of the peak accelerating field — through a linac produces a negative

position-momentum (z, pz ) correlation, or longitudinal chirp, on the beam,

which causes it to be compressed in the chicane, as the chicane path length

of higher momentum electrons is shorter than that of lower momentum par-

ticles. This process is described in more detail in the case of the Neptune

compressor in Chapter 1, and more of the technical details associated with

the operation of this device are given below.

The use of a chicane to compress the beam has a price, however, paid in

the deterioration of the transverse phase space and energy distribution of

the beam. These effects arise from the collective fields of the beam, which

may act directly through transverse forces, or indirectly where a longitudinal

force changes the energy of the electron during the bend process, and the

electron trajectory attains an additional error in position through momen-

tum dispersion mismatch. When the beam is ultra-relativistic (γ  1), the

163
energy changes induced during the motion are expected to arise mainly from

coherent synchrotron radiation (CSR) [83]. These energy changes may be so

pronounced that a new type of microbunching instability may develop [84].

Previous studies of collective effects during the chicane compression pro-

cess have been carried out at the European Organization for Nuclear Research

(CERN), where transverse emittance growth [85] and changes in the momen-

tum spectrum [86] were observed in the compressed beam. These studies

were performed in the energy range of 40-60 MeV, and the observed emit-

tance increase and momentum spectra were compared to predictions of the

simulation code trafic4 [87]. From this comparison, evidence for strong CSR

emission was deduced, and the implication is that the acceleration fields of

the bending electrons provided the dominant collective effect in the experi-

ment. Other measurements of magnetic compression have been performed

at higher energies [88], with CSR again being the dominant mechanism caus-

ing the observed emittance growth.

While these initial measurements were a first test of collective effects in

high brightness beam compression, the distortion of the transverse phase

space was quantified only through examination of the normalized rms emit-

tance, εn,x = βγ x 2 x
2  − xx
2 . In the experiments performed at Nep-

tune and described here, the horizontal trace space of the beam is sampled

directly using the slit based system described in Chapter 3 (the dimensions of

the collimating masks were in fact identical in these two experiments). This

allows an accurate, shot-by-shot reconstruction of the trace space, free from

the strong (especially after compression, since, as will be shown, the beam

current increases by a factor of five) space-charge forces that dominate the

164
beam’s expansion in a drift space. Because the measurements were single-

shot, other parameters important in the analysis of this experiment, such

as charge and PWT phase were obtained simultaneously. In addition, the

Neptune measurements were performed at a lower beam energy (below 12

MeV), and thus explore a regime where velocity fields may play a dominant

role. Such lower energy compression may be needed for experimental ap-

plications, such as Thomson scattering [75], which demand more moderate

(20-50 MeV) final energies.

The measurements presented in this chapter were performed using the

Neptune photoinjector as shown in Figure 1.8. The results of these measure-

ments show a strong bifurcation in the trace space of the compressed beam,

in which there can be seen two distinct species of beam particles, overlap-

ping in x but separated in x
. This effect was observed to be correlated to

the folding of the beam distribution in (z, x) space that occurs in the chicane

compressor, and was analyzed using several different computational mod-

els. The results of the simulations of the Neptune experiment indicate that

space-charge fields (that is, the components of the beam’s transverse velocity

fields which apply force according to Fx ∝ 1/γ 2 , found in Chapter 2) play a

major role in the dynamics of the compressor. To understand the effect that

space-charge forces have on the evolution of the beam’s transverse phase

space, a heuristic model was developed which constructs the beam from a

series of longitudinal slices (just as in the emittance compensation model),

and examines how these slice interact though space-charge forces. A simu-

lation code, termed bender, was written to apply this model to the case of

the Neptune experiment, and the results of these simulations show behavior

165
consistent with aspects of the experimental data, including the phase space

bifurcation.

4.1 Experimental Measurements at Neptune

The basic operating principle of the compression system employed at Nep-

tune was outlined in Section 1.3.5, and has been discussed previously by

others [4, 39]. Here some of the finer points of running this system are

described. Making the compressor work was, of course, the required first

step towards making the beam measurements described below. That does

not make this step a trivial one, however, as it required a fair amount of at-

tention to detail. Furthermore, the adjustments of the system parameters

required for its proper operation had consequences on beam dynamics and

in fact accentuated the phase space distortion that we observed. These ad-

justments were made on the relative strength of the fields in the dipoles, and

were made necessary by two non-ideal aspects of the system: misalignment

of the dipoles, and the three-dimensional nature of the dipoles (which caused

significant fringe fields).

The chicane uses 11.3◦ edge angles at the entrance of the first and exit

of the last dipoles, as shown in Figure 1.13. These angles were implemented

in the design of the compressor to provide (horizontal) focusing in the bend

plane, and vertical defocusing, to mitigate the strong vertical focusing of the

rest of the system. This vertical focusing would cause a ballistic waist in the

beam size inside the chicane if not weakened, and therefore, the entrance

and exit angles are an important feature of the magnet design. This feature

166
also effects the horizontal steering of the beam. This can be seen from in-

vestigation of Figure 1.13, since the length of the magnet now depends on


the horizontal position of a particle’s entrance, l = l0 + ∆x tan θedge , where

l0 is the magnet’s length for the design particle and ∆x is the horizontal

displacement from that design position. Therefore, by Equation (1.34), posi-

tional errors in the beam at the chicane entrance lead to trajectory errors in

the system and it can be shown through linear transport matrix analysis [4]

that the beam exiting the chicane will be displaced in both x and x
. If the

parameters for the Neptune case are applied to the linear analysis, Rbend = 31

cm, then a 2 mm offset in x at the chicane entrance gives a 1 mm offset at the

exit (this is consistent with the focusing properties of the system) and a 2.4

mrad angular offset. This angular error increases as the radius of curvature

decreases (i.e., the field in the magnets increases or the energy of the beam

decreases, as it does when the PWT is run off-cress for compression) and is

clearly noticeable at a view screen downstream of the chicane. In fact this

steering error was first discovered by observing the motion of the beam at

a downstream view screen as the field in the magnets was increased. The

observed motion of the beam centroid prompted a careful measurement of

the dipole alignment and an error of approximately 2 mm was found in the

+x direction (which effectively lengthens the first and last magnets).

The other factor that contributes to the effective length of the first and last

magnets is the fringe fields that arise due to the finite gap size in comparison

to the pole face area (in other words, the fields produced by these magnets are

three dimensional, and not well modeled by 2-D computer codes). In order to

determine the extent of the fringe fields present in the system, the magnets

167
 1.5

0.5
By [kGauss]

-0.5

-1

-1.5
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

z [m]

Figure 4.1: Magnetic field in the chicane computed by amperes. The max-
imum field amplitude produces a 31 cm radius of curvature for a 11.8 MeV
beam.

were modeled in the three dimensional magnetostatic code, amperes [89].

The gap field given by simulation is plotted versus longitudinal position in

Figure 4.1. The important feature of this plot is that the fringe fields of the

first and last magnets extend a significant distance beyond the end of the

chicane. Thus, at the exit of the chicane the last effects of the magnet (which

should perfectly remove any dispersion in the beam, in the absence of space-

charge) are applied slowly, giving a greater distance over which space-charge

forces act to diminish the beam quality.

The combination of these two factors resulted in the effective lengthening

168
of the first and last magnets, and thus an increased bend angle in the first

magnet which altered the beam trajectory in the rest of the chicane and gave

a non-zero angle and non-zero dispersion at the exit of the system. This

problem was solved in the experiment by slightly increasing the strength of

the magnetic fields in the middle dipoles, in such a way that the bend angle

for each magnet was the same.

This fix was made possible by small trim coils, which were wrapped around

each of the main coils of the four dipole magnets. In this set-up, the main

coils were wired in series with a sufficiently high power (about 250 Watt) cur-

rent supply to produce the main fields in the chicane. The trim coils were

then connected to individual power supplies to allow minor (< 10%) adjust-

ments to the field in each magnet. Before installation of the dipole magnets

into the Neptune beamline, the fields in each magnet were measured as a

function of both main and trim current using a Gauss-meter. These mea-

surements allowed us to control the ratios of the fields in each magnet for a

given main current, using the current of each trim coil.

The same procedure that originally alerted us to this problem was used

to correct it. That is, the beam position was observed on the view screen

approximately one meter downstream of the chicane exit as a function of

current. The fields were adjusted using the trim coils to produce equal fields

in the middle magnets which were a given fraction higher than the fields in

the outer magnets. This increment was determined experimentally to be 3%,

as this was the amount that minimized steering as a function of main coil

current (the final steering of the system was measured to be a small fraction

of the beam size on the screen, as compared to the uncompensated system,

169
Figure 4.2: LabVIEW program controlling the main and trim coil settings of
the chicane. The “Auto-trim” button sets the trim coil strengths to adjust the
fields in the four dipoles to eliminate dispersion after the chicane.

which walked the beam across the full width of the YAG crystal, about 1

cm), and therefore, made the bend angle equal in all four magnets. This

proceedure also served to eliminate the residual momentum dispersion after

the chicane.

Once the proper field balance was determined, the values of the trim coil

currents required to achieve this balance were programed-in to the LabVIEW

based control system. This is illustrated in Figure 4.2, which shows the Lab-

VIEW virtual instrument (VI) written to control the chicane. Here the proce-

dure to de-Gauss the magnets is also automated, which was required since

hysteresis would cause the fields to vary significantly compared to the 1%

precision required to kill dispersion. These automations illustrate the utility

of this type of control system, as they allowed us to perform the experiments

170
in a minimum amount of time, which is an important consideration in the

multi-shot measurements described below.

4.1.1 Pulse Length Measurements

The next step in this study was to test the ability of our system to compress

the electron beam. This required a direct measurement of the beam’s bunch

length which when compressed, is less than one pico-second (rms). The mea-

surement of sub-picosecond electron beams has been a topic of much recent

interest [] (as one might guess, given the bunch lengths required by advanced

photoinjector applications), since these measurements are at or beyond the

capability of conventional diagnostics, such as streak cameras. For that rea-

son, an alternative measurement technique using coherent transition radia-

tion, emitted by the beam upon impact with a thin foil, has been developed by

several labs [53, 54, 90], and is used in the diagnostic employed at Neptune

as well.

The process of a charged particle crossing the boundary of a perfect con-

ductor can be modeled as a collision of that particle with its image charge.

The radiation produced by this annihilation process is termed transition ra-

diation. The angular distribution of this radiation has been found to be inde-

pendent of frequency (in the limit that the metallic boundary acts as a perfect

conductor) and directed in a narrow cone with angular width is parameter-

ized by 1/γ, a characteristic common to many types of radiation produced

by relativistic particles. The radiation produced by the electron bunch then,

is found by summing the radiation fields of all of the beam particles. For

171
 e-beam CTR foil

collecting lens
45∞-polarizer
1

7 2
detector 6 reference
detector
3
90∞-polarizer
4

5
translator retro-reflectors

Figure 4.3: Schematic of the CTR-based bunch length diagnostic.

wavelengths that are long compared to the length of the bunch, the total

field created by the beam is coherent (∝ N), and therefore the total energy is

given by N 2 times the energy radiated by a single electron. As the radiation

wavelengths approach the size of the beam, this changes and, as argued in

Ref. [54], the total spectral energy radiated by the beam is written as

2
Ẽ (ω) = 2π Ẽ1 N 2 |ρ̃ (ω)| , (4.1)

were Ẽ1 is the single particle spectral energy, and ρ̃ (ω) is the Fourier trans-

form of the bunch longitudinal profile.

The CTR-based diagnostic used in this experiment and shown schemati-

cally in Figure 4.3, is a polarizing Michelson interferometer based on beam

splitters which use a transmission wire grid with a 100 µm wire spacing. This

spacing sets a lower limit on the wavelengths measurable by this device. The

172
CTR in this system is generated by colliding the beam with an aluminum foil

mounted at 45◦ with respect to the beam axis, leaves the vacuum through

a quartz window, and is focused and directed toward the interferometer by

an off-axis paraboloid mirror. This light is then split into two parts by the

first wire grid, which is oriented at 45◦ from vertical. One of these parts is

sent to the reference detector. This reference signal is needed to normal-

ize the interferogram signal, since this a multi-shot measurement which is

very sensitive to beam charge fluctuations (E ∝ N 2 ). The other half of the

light is split again by the 90◦ polarizer, and the horizontally and vertically

polarized components are sent to two different retroreflectors, one of which

mounted on a translator, allowing the path length to be varied between the

two arms. The two polarizations are recombined at the 90◦ polarizer with a

phase shift controlled by the path length difference between the two arms.

The portion of this that is reflected by the 45◦ polarizer into the detector

(path 7 in Figure 4.3) obeys the relation

1  
E7 (ω) ∝ √ eiωt 1 − eiωτ . (4.2)
2

Thus, the signal that reaches the second detector is the autocorrelation of

the CTR signal, which is in turn, is the autocorrelation of the beam’s temporal

profile, since the Convolution Theorem [91] relates the time and frequency

domains in this case by


ρ (τ + t) ρ (τ) dτ ⇐⇒ |ρ̃ (ω)|2 , (4.3)

where the Fourier transform and inverse transform are used to go from one

quantity to the other.

173
 In practice, the situation is complicated by the behavior of the longer

wavelengths in the diagnostic. For radiation wavelengths more than a few

mm, which is highly coherent and carries a significant part of the total energy,

the interferometer acts as a high pass filter. This effect is due to the diffrac-

tion of these wavelengths beyond the physical apertures associated with the

optics and detectors in the device. This lack of low frequency components

in the detector signal makes the Fourier transform of the data difficult to

interpret. Instead, a time domain fitting approach was developed to analyze

this type of data in Ref. [54], and was used in the Neptune experiment. The

utility of this approach comes from the observation that the detected signal

is in fact the filtered autocorrelation of the beam distribution. The frequency

dependence of this filtering is taken by this model to be

2ω2
g (ω) = 1 − e−ξ , (4.4)

which has the proper long wavelength attenuation, and a frequency cut-off at

ξ −1 . This filter function could be derived by the aperturing of a diffraction-

limited transverse Gaussian mode photon beam with an initially uniform

frequency spectrum in the far field, a scenario that bears resemblance our

present case. Assuming this filter, the measured spectral beam density is

given by

ρ̃meas (ω) = ρ̃ (ω) g (ω) , (4.5)

and the spectrum of the measured signal is

% 2 2 2 2
&
s̃ (ω) = |ρ̃meas (ω)|2 = |ρ̃ (ω)|2 1 − 2e−ξ ω + e−2ξ ω . (4.6)

If we assume the longitudinal beam distribution to be Gaussian, then the

Fourier transform of s̃ (ω) gives an expression for the form of the signal in

174
the time domain,
 
(τ−τ0 )2 (τ−τ0 )2
(τ−τ0 )2 2σ − σ −
s (τ) ∝ e ).
−
4σ 2 − e ( 4 σ 2 +ξ 2 )+ e ( 4 σ 2 +2ξ 2
(4.7)
σ +ξ
2 2 σ2 + 2ξ 2

In the measurements shown below, Equation (4.7) was used to fit the in-

terferometer data to obtain the rms pulse length σ . As a part of the fit the

cut-off parameter ξ, which is a property of the diagnostic, was calculated for

each measurement. This parameter was consistently found to be about 3

psec, corresponding to a filter cut-off normalized wavelength of 1 mm.

The procedure used to measure the bunch length was relatively straight

forward. The chicane was first run in spectrometer mode (that is, the first

two magnets were turned off, and the last two were set to bend the beam

in the same direction). The magnet currents and PWT phase were set to put

the beam on the view screen at the end of the spectrometer arm, with a mini-

mum energy spread. This minimum energy spread does not occur at the peak

energy, but roughly 5◦ ahead of crest, which compensates for longitudinal

space-charge effects (occurring mostly before the PWT). This procedure de-

termines the chicane current needed to produce the design bend radius when

running (almost) on-crest in the PWT (this corresponded to a beam energy of

11.8 MeV). The PWT phase required to minimize the energy spread was mea-

sured using the rf mixing scheme described in Chapter 1, and recorded as a

reference for the phase adjustments described below.

At this point the chicane was set back to compressor mode, using the main

coil current setting found in the spectrometer. The trim coils were adjusted

to produce the same bend angle in each magnet, as described above, and the

CTR reference signal was observed as the PWT phase was varied. This was

175
done to find the injection phase which produced the peak compression of the

beam, since it can be shown [92] that the number of CTR photons produced

is inversely proportional to the bunch length of the beam. It is important

to point out here that the CTR signal is proportional to N 2 , so charge fluc-

tuations needed to be averaged out by taking 10-15 detector measurements

for each PWT phase. This maximum compressing phase was consistently

found to be approximately 25◦ ahead of the minimum energy spread phase,

or about 30◦ ahead of crest.

Next the phase for maximum compression was recorded, to be used both

as a check that the PWT phase does not drift by more than one or two degrees

in the time required for the interferometer scan, and to reset the PWT to this

phase between scans at different magnetic field strengths. The interferome-

ter scans themselves were automated using a LabVIEW program, which mea-

sured via communication with oscilloscopes the peak signals from both CTR

detectors, and the PWT phase for each shot. The phase measurements were

important as they allowed further verification that the phase drifts in the

measurement were acceptably small, and the filtering of “bad” data points,

such as those occurring in shots where the beam was lost due to rf break-

downs. The use of LabVIEW to automate the measurement procedure was

important in this case, because the amount of data recorded is large and at

the same time the measurement should be completed as quickly as possible,

to minimize drifts in the system.

Finally, the interferometer measurement was performed at a constant

(maximum compressing with a 5 amp dipole current) phase and with vary-

ing magnet currents. The results of two of these measurement is given in

176
Figure 4.4, which shows the detector signal data points, normalized by the

reference detector signal and averaged over 15 beam shots, versus delay arm

length (measured in psec). Part (b) of the figure gives the interferometer data

at maximum compression which, when fit to Equation (4.7), yields a com-

pressed pulse length of 0.6 picoseconds rms. In the data shown in part (a),

the dipole current (and therefore bend angle) is cut in half, resulting in less

compression, and a measured rms pulse length of 2.7 picoseconds. Notice

that the data for the fully compressed case has much more noise in the tails

of the filtered autocorrelation than the slightly compressing case, as the com-

pression process and the CTR signal are more sensitive to phase drift and

shot-to-shot fluctuations at smaller pulse lengths. It may also be an indica-

tion of high frequency structure in the compressed beam.

This pulse length data is plotted in Figure 4.5, along with the measured

pulse length with the chicane magnets off, and the PWT set for minimum

energy spread. Unfortunately, this longest pulse length data point could not

be found using the interferometer. The signal in this case was too weak to

perform the measurement due to the strong filtering of wavelengths above

ξ = 3 psec in the diagnostic. This point was obtained instead, by measuring

the beam momentum spread. Using Equation (1.39), the momentum spread

gives the approximate pulse length by

σδp (krf σz )2
≈ . (4.8)
p0 2

This approximation is fairly rough, as it does not consider space-charge and

other effects on the longitudinal distribution; however, the result agrees with

both streak camera measurements of the drive laser and parmela simula-

177
 1.2

1.0

Normalized Detector Signal 0.8

σt = 2.7 psec
0.6

0.4

0.2 Detector Signal

(a) Time Domain Fit
0.0
0 5 10 15 20 25
Delay [psec]

1.2

1
Normalized Detector Signal

0.8

0.6
σt = 0.64 psec
0.4
Detector Signal
Time Domain Fit
0.2 (b)

0
0 5 10 15 20
Delay [psec]

Figure 4.4: Interferometer data plotted with the time domain fit function of
Equation (4.7), used to find the pulse length for (a) 2.5 amp chicane current
and (b) 5 amp current, corresponding to maximum compression.

178
 5

tredi Simulation
4 Data

σt [psec]
3

2
energy spread
measurement
1

0
-0.2 0 0.2 0.4 0.6 0.8 1 1.2

B/Bmax

Figure 4.5: The measured rms bunch length plotted versus the chicane mag-
netic field normalized to that required to produce the design bend radius (1.2
kGauss).

tions of the system.

For comparison, Figure 4.5 also shows the pulse lengths obtained by mod-

eling the system with the simulation code tredi [93] (this code will be de-

scribed below in the analysis of the emittance measurements). The agreement

between data and simulation is excellent, which gave us confidence both in

the proper operation of the compressor, and in our knowledge of the system

parameters. Both of these are required for the emittance measurements and

analysis described below.

179
4.1.2 Emittance Measurements

At this point the emittance of the beam was measured using the slit system

described in Chapter 3 for different chicane current and PWT phases. In the

measurements described below, the initial steps were the same as those used

in the pulse length measurements described above. Namely, the phases of

minimum energy spread and maximum compression were found using the

spectrometer arm and the CTR detectors, respectively. At the same time,

the chicane current required to bend the minimum energy spread (nearly

maximum energy) beam through the design bend angle was also found, thus

bracketing the range of chicane currents and PWT phases to be used in com-

pressing the beam to various degrees, and measuring the emittance.

4.1.2.1 Slit Image Analysis with Bifurcations

The slit data discussed below shows a great deal of structure. The intensity

profile of a single beamlet is not a simple Gaussian, or other well-defined,

singly peaked distribution, but has clearly distinguishable features. These

features complicate the analysis of the slit-collimated beam images, as we

wish to measure both the rms emittance and reconstruct the trace space of

the beam. The emittance measurement requires the calculation of the rms

size of the beamlet on the detecting screen, while the trace space reconstruc-

tion requires that the structure of the beamlet be preserved.

In order to simultaneously account for the beamlet structure and compute

180
rms beamlet sizes, the beamlet intensity profiles are fit to a sum of Gaussians,

2 
 x − x
Aj exp − .
j
Ibeamlet = 2 (4.9)
j
2σ j

The weight of the jth peak is then Aj σj and the mean square size of the ith

beamlet is given by
) %
2 &
A j σj xj − x + σj2
  j
xi2 = ) . (4.10)
A j σj
j

This beamlet fitting method is demonstrated in Figure 4.6 for an typical slit

image found in these measurements. In this figure, the intensity of the sec-

ond beamlet is fitted using j = 3. Notice that there are in fact two distinct

parts of this beamlet, as seen in both the image and the intensity plot. The

last Gaussian accounts for the shape of the intensity distribution in the tail

of the first peak, and improves the fit to the data. When computing the beam

size, this fitting method has an advantage over a direct rms measurement [as

done with Equation (3.5)] in that it is immune to noise in the tails of the dis-

tribution, which contribute heavily to the direct rms measurement and may

result in an overestimate of the beam size in that case. The fit function of the

beamlet is then used in the beam’s trace space reconstruction, as it defines

the x
distribution of the beamlet at a location in x defined by the geometry

of the slits. This detailed reconstruction would not be possible with a simple

rms beamlet width analysis.

4.1.2.2 Emittance and Trace Space with Compression

With the ability to compress the beam with the chicane, measure its length

with the CTR diagnostic, and analyze the slit images, we then proceeded to

181
 (a)

5000
Beamlet Intensity [Arbitrary Units]

Data
Peak #1
4000 Peak #2
Peak #3
Fit

3000

(b)
2000

1000

0
60 70 80 90 100 110

Pixel Number

Figure 4.6: (a) An image of a partially compressed beam collimated by the

emittance slits. The width of this slit image is 7.6 mm. (b) The (vertically
summed) intensity profile of the second slit in (a). The data is fit in this case
with three Gaussians to compute the rms size of the beamlet and retain the
phase space structure.

182
measure the horizontal emittance of the electron beam. It should be noted

here that the vertical emittance could, of course, not be measured with the

(vertically oriented) slit-mask. This is not a concern for us since the bend

plane of the chicane is horizontal and therefore, the manipulation of the

beam distribution is limited to the x and z phase planes. The vertical beam

size in the chicane effects (through the beam density) the processes that

lead to emittance growth in the x dimension, but these processes do not

effect the beam in the y dimension. This fact was verified experimentally,

by performing vertical and horizontal quad scans for the compressed and

uncompressed beam. The emittance computed by these scans was not a

completely accurate number, as the beam is highly space-charge dominated

in this system, leading to the quad scan difficulties discussed in Chapter 3,

although not at near the level encountered in the LLNL experiments. The

comparison of the quad scans in the compressed and uncompressed cases

was informative, however, as the results in the y dimension were roughly

the same in either case, while there was a drastic increase in the x-emittance

when the beam was compressed.

The first emittance measurements were taken as a function of PWT phase,

with the chicane current fixed at that required to give the design bend angle

for an 11.8 MeV (minimum energy spread) beam. The results of this measure-

ment are shown in Figure 4.7. The figure shows the normalized emittance

increasing from about 8 mm mrad in the uncompressed case to approxi-

mately 20 mm mrad when the beam is fully compressed. The bunch length

increases very close to linearly with the PWT phase in the range plotted in the

figure with a phase of 60◦ corresponding to the minimum 0.6 psec bunch and

183
 25

20

εx,n [mm mrad]

15

10

5
55 60 65 70 75 80 85 90

Linac Phase [◦ ]

Figure 4.7: The horizontal normalized emittance of the beam compressed in

the chicane by varying the PWT phase. The pulse length decreases approxi-
mately linearly from 4 psec (rms) at φ = 85◦ to 0.6 psec (rms) at φ = 60◦ , as
the PWT is run progressively further off-crest.

85◦ corresponding to the uncompressed length of 4 psec. Notice that there

is a sharp change in the emittance at phase φ = 72◦ . This change marks the

phase at which the beamlets begin to show structure, as shown in Figure 4.8

below.

The fact that the emittance increases as the beam is compressed is an ex-

pected result since, as mentioned above, similar behavior has been observed

in Refs. [85] and [86]. On the other hand, the trace space reconstructions,

shown in Figure 4.8, do yield unexpected structure. In this figure the slit im-

184
ages and corresponding trace spaces are shown for four points in the emit-

tance plot of Figure 4.7. Here we see that as the emittance increases, the

beam bifurcates in trace space into two separate parts. In can be seen that

this is so from the slit images where, as the compression increases, the multi-

peaked beamlet intensities appear. Two separate peaks observed at the view

screen, which originate at the same slit, imply that the phase space of the

beam at that slit consisted of two parts, with the same position but differing

momentum.

The multi-peaked beamlets first appear in the data at the same phase as

the sharp emittance increase and are in fact, the source of this rise. This

bifurcation was a consistent feature of the data collected in this experiment,

and as Figure 4.8 shows, is strongly dependent on the degree of compression.

Note also from the images that the separation of “sub-beamlets” arising from

one slit is a function of vertical position, with the strongest separation oc-

curring in the middle (most dense part) of the beam.

In a different run, the emittance was measured again, but in this case

versus phase (at constant chicane current) and chicane current (at constant

phase). The emittance found in these measurements is shown in Figure 4.9.

Here the emittance is plotted versus pulse length directly. The figure shows

behavior similar to that of Figure 4.7, and the emittance growth as well as

the phase space bifurcation was observed to increase as the beam was com-

pressed by changing either parameter.

185
 (a1) (b1) (c1) (d1)

186
(a2) (b2) (c2) (d2)

Figure 4.8: Slit-mask images and traces space reconstructions used to calculated the emittance of four of the
points in Figure 4.7, going from (a) no compression to (d) maximum compression.
 25

Changing Current
Changing Phase
20
εx,n [mm mrad]

15

10

5
0.5 1 1.5 2 2.5 3 3.5 4 4.5

σt [psec]

Figure 4.9: Comparison the emittances obtained by varying the PWT phase
at a constant chicane current and varying the chicane current at a constant
PWT phase.

4.1.2.3 Emittance Versus Beam Size

The observation that the trace space bifurcation is a function of the beam’s

density prompted the investigation of emittance growth as a function of the

spot size at the chicane entrance. In this measurement the beam size at

the entrance to the chicane was varied using the emittance compensating

solenoid. The solenoid was chosen as the size control “knob” because it,

combined with the focusing of the PWT, was capable of producing the widest

187
range of input spot sizes. The alternative choice was to use the quadrupole

triplet located between the PWT and the chicane (see Figure 1.8), but this

option does not significantly change the beam size at the chicane entrance,

both because this triplet is located too close to the chicane for this purpose,

and because the beam at this location is small (after being accelerated near

the invariant envelope in a high gradient linac) and therefore, not easily fo-

cused further. The disadvantage of using the solenoid is that it alters the

emittance compensation process, and the emittance going into the chicane.

It was therefore necessary in this measurement to find the emittance without

compression (chicane current off, but PWT phased 30◦ ahead of crest), as well

as with compression, for different solenoid settings.

The emittance contributions from the linac and the compression process

add in squares, so the emittance growth from the chicane is extracted from

the data as

2 2
εchicane = εtotal − εlinac . (4.11)

The results of this measurement as given by Figure 4.10, which shows that

there was in fact a strong correlation between the beam size in the chicane

and the emittance growth that resulted. The amount of emittance growth

was again associated with the level of phase space distortion and bifurcation

observed in the slit images, with the smaller beam showing a bifurcation

that is both larger, and more strongly dependent on vertical position within

the beam. This is shown in Figure 4.11, which gives the slit images used to

calculate the emittances in Figure 4.10b at the 217 and 202 Amp solenoid

current data points, respectively.

188
 35

30

εx,n,chicane [mm mrad]

25

20

15

10

0
1 1.2 1.4 1.6 1.8 2 2.2 2.4

σx [mm]

Figure 4.10: The emittance growth resulting from compression in the chicane
plotted versus the horizontal rms size of the beam entering the chicane.

In summary, the slit-based emittance measurements of the beam com-

pressed by the chicane show an emittance increase from approximately 7

mm mrad at the 4 psec pulse length of the beam produced by the linac, to a

number measured to be as high as 30 mm mrad (which is strongly dependent

on the beam size in the compressor) at the fully compressed pulse length of

0.6 psec. This emittance growth is a strong function of the amount of com-

pression, but not of the magnet bend angle or PWT phase alone, as similar

results are obtained by compression varying one parameter while holding

the other constant. In horizontal trace space, the beam bifurcates into two

components which are overlapping in x but offset in x
. In the measurement

where the compression was increased by scanning the PWT phase, the emit-

tance was found to increase sharply with phases further than 28◦ ahead of

189
 (b)
(a)

Figure 4.11: (a) Slit image used to compute the emittance growth for the 217
Amp solenoid current point in Figure 4.10b. (b) Slit image used to compute
the emittance growth for the 202 Amp solenoid current point in Figure 4.10b.

crest. This was also the point where phase space bifurcation appeared. The

bifurcation and emittance growth were found to be a function of beam den-

sity, and were greater when a smaller beam was injected into the compressor.

4.2 Analysis

The collective effects that may be significant to the process of emittance

growth in the compressor can be separated into two categories: effects due

to the beam’s acceleration fields, and those due to the beam’s velocity fields.

The major acceleration field effect, as mentioned above, is CSR, which has

been considered to be the most important emittance growth mechanism for

experiments at higher energy. In the velocity fields category there is space-

190
charge in different varieties: the “standard” space-charge force, which obeys

the usual 1/γ 2 dependence, and centrifugal [94] and non-inertial [95] space-

charge forces, which do not. The importance of these non-“standard” forces

that occur in bends is not clear, and it has been asserted [96, 97] that they

may be largely cancelled through potential energy considerations.

4.2.1 Simulation

In this experiment the relatively low energy (γ ≈ 21) forces us to consider

space-charge (the 1/γ 2 variety) as a possible source of the effects observed.

In order to unfold the various physical effects involved in this experiment,

several different codes were employed to simulate the evolution of the beam

in the chicane. As the physical processes modeled by these codes are differ-

ent, we gain information about the sources of observed emittance growth in

the system by comparing the simulations to the measured behavior of the

beam.

The first code used was parmela, both to provide input phase space dis-

tributions for other codes, and to run through the chicane using the three

dimensional point-by-point space-charge calculation. This calculation uses

a “quasi-static” approximation to compute the space-charge fields. In this

approximation the beam is transformed to its rest frame where the static

electric fields are used to calculate the forces on the particles. The num-

ber of simulation particles used in these simulations was limited to 2500

for runs which are compared to tredi, and 10,000 for more detailed stud-

ies. This relatively low number was necessary because both 3-D parmela

191
and tredi (which is also 3-D) are N 2 computations, which become very slow

as this number is increased further. To minimize the noise inherent in this

low number of simulation points, the parmela simulations from the cath-

ode to the entrance of the chicane (where the beam is always axisymmetric)

were done using the conventional r –z mesh space-charge algorithm, and the

phase space distributions produced by these runs were restarted in the 3-D

space-charge mode. It is important to understand that the only collective

effect modeled by parmela in these simulations is (1/γ 2 ) space-charge; the

effects of radiation fields and non-inertial space-charge are not included.

The second code used to model the beam in the chicane was tredi, which

computes the electromagnetic fields of the beam on a three dimensional mesh

using the Lienard-Wiechert retarded potentials [93]. In order to do this, the

code must keep track of the history of the particle’s position, r (t), which

severely limits the number of macroparticles that can be used due to mem-

ory requirements. By computing the fields in this manner, the effects of

both acceleration and velocity fields are implicitly included in the calcula-

tion. Thus, the difference between tredi and parmela simulations should

indicate the effects of CSR and non-inertial space-charge.

These codes were used to simulate the phase scan data given in Figure 4.7.

The results of these simulated scans are given with the experimental data

in Figure 4.12. As one can see, the agreement between the data and the

two simulation codes is quite good. This agreement suggests that the domi-

nant emittance growth mechanism in this experiment was space-charge. To

double-check this implication, the code elegant [98] was run for the max-

imally compressing PWT phase. Elegant is a linear matrix based particle

192
 25 Emittance data
parmela simulation
tredi simulation
20 elegant simulation
εx,n [mm mrad]
15

10

0
55 60 65 70 75 80 85 90
PWT Phase [◦ ]

Figure 4.12: Results of the parmela and tredi simulations of the beam
after compression plotted versus PWT phase. The experimental data from
Figure 4.7 is plotted as well for comparison. The single elegant data point
at the phase for maximum compression shows the effect of CSR alone.

tracking code that calculates the energy loss and transverse kicks on the par-

ticles due to CSR in a one dimensional approximate model. This simulation

point is shown as well in Figure 4.12. The comparatively small (< 1 mm

mrad) emittance growth predicted by elegant indicates that, while CSR may

have a small effect in this system, it does not account for the majority of the

observed emittance growth. This fact, combined with the agreement between

parmela, tredi, and the experimental data provide strong support for the

notion that space-charge is the dominant physical effect in this experiment.

In addition, these simulations also show distribution bifurcations and beam

193
size dependence, which will be shown below, and can be understood through

a heuristic model of the space-charge forces in the compressed beam.

4.2.2 A Heuristic Model

The role space-charge plays in this process can be understood by applying

a simple model of the beam in propagating it through the compressor. This

model is the same one used in the emittance compensation analysis of Chap-

ter 2. Specifically, the beam is broken into longitudinal slices, and the dif-

ferences in the evolution of these slices can be observed to investigate the

effect of space-charge. Unlike the rectilinear motion this model was applied

to in Chapter 2, the beam slices in the chicane will change longitudinal posi-

tion with respect to each other, and therefore we must consider transverse

space-charge forces between different slices.


This model has an advantage in the handling of dispersion [∆x ∆(δp/p)]

in the chicane. The parmela simulation of the beam entering the chicane in

Figure 4.13 shows that the beam’s longitudinal phase space is highly corre-

lated, as required for a large compression ratio. On the other hand, the un-

correlated energy spread within a longitudinal slice is very small. This fact

allows us to model the individual slices as monoenergetic, and the beam’s

dispersion through the chicane is given by the relative motion of different

slices.

To visualize the phase space “gymnastics” the beam performs in the chi-

cane, consider the motion of the slice centroids using this model. In Fig-

ure 4.14 the longitudinal phase space and the (z, x) configuration space of

194
 22.5

22

21.5

βz γ
21

20.5

20

19.5
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3

δz [cm]

Figure 4.13: parmela simulation of the beam’s longitudinal phase space at

the entrance to the chicane.

these centroids is given at three positions along the chicane. The slice motion

in this case is calculated without space-charge forces, using the parmela gen-

erated initial phase space shown in Figure 4.13. The beam’s motion in phase

space is easily understood from an analogy to transverse focusing. The cor-

relation between position and momentum created by the PWT is removed in

the chicane, which acts as a longitudinal drift. The distribution in longitu-

dinal phase space then, essentially rotates to an upright position, and the

minimum pulse length, which is determined by the rf curvature in the PWT,

is achieved. Notice that in the last set of phase and configuration space plots

in Figure 4.14, this phase space rotation is completed in the middle of the last

magnet (the pulse length actually increases slightly at the end of the magnet).

195
 x x x

z z z

δp δp δp
p p p

z z z

Figure 4.14: Motion of the beam slice centroids in the chicane. The input
phase space is given by a parmela simulation, and the slices are propagated
without space-charge.

196
 5

ηx [cm]
3

0
0 1 2 3 4

s
Rbend θbend

Figure 4.15: Dispersion in the chicane. This calculation is simplified by ignor-

ing the small drift spaces between the magnets. Note that while ηx is greatest
in magnets 2 and 3, ηx /R is negative here because of a sign convention in

R. Therefore the chicane’s path length parameter αc = 4Rbend1 θbend ηRx ds is
negative, as needed for compression.

In configuration space we see that as the beam compresses, its motion

in x is controlled by dispersion in the system. The dispersion function

∂x
ηx (s) ≡ ∂ (δp/p )
, in the chicane is shown in Figure 4.15. The configuration

of the magnets requires that the dispersion be symmetric about the middle

of the compressor, where the dispersion is maximized. Likewise, symmetry

requires that the dispersion (and its derivative) are killed at the end of the

system. Thus, the function of the third and fourth magnets (aside from com-

pressing in z) is to remove the dispersion created in the first two. In fact,

this is the most significant function of the last magnet, since the beam is fully

197
compressed by the first three. The rf curvature induced “comma” shape in

the longitudinal phase space also shows up in the beam configuration space

in the last magnet. The difference here is that in order to remove dispersion,

this comma shape in configuration space is squeezed down onto itself. In

other words, the result of this process is a beam which is folded into two

parts, and these parts are forced on top of each other in the last chicane

magnet.

As one might guess, the act of folding these two portions of the beam

together is complicated significantly when space-charge forces are included

in the problem. Therefore, to proceed with this model we must specify the

distribution of the slice in order to determine its space-charge forces. We

chose for this model to consider the slices to be uniformly filled ellipsoids

of charge with density defined by




 x2 y2 z2

 ρ0 : a2
+ b2
+ c2
<1
ρ x, y, z = , (4.12)



0 : otherwise

where a, b, and c are the semi-major axes of the ellipsoid in the x, y, and

z dimensions, respectively. This type of beam distribution has been used

extensively in charged particle beam physics because it is well known [99,

100] that the fields inside the beam are both linear and a function of only the

dimension of the field component. For example, Ex is written as

Ex (x) = α0 x, (4.13)

where α0 is a constant determined by the values of a, b, and c, and of course,

the total charge of the slice. This constant has been approximated for various

special cases [100], and its most general form is given in Appendix A.

198
 Not as well know to the beam physics community is the expression for

the electric fields outside the ellipsoid, which take a similar form. Here Ex

becomes [101, 102]


Ex x, y, z = α (λ) x, (4.14)

where α is now a function of the confocal ellipsoidal coordinate λ, defined by

the generating function in this coordinate system [103]

x2 y2 z2
f (s) = + + = 0. (4.15)
a2 − s b2 − s c2 − s

The roots of the equation f (s) = 0 define the three ellipsoidal coordinates λ,

µ, and ν, which, in a rough analogy to spherical coordinates, can be thought

of as the radial, and two angular variables, respectively. The surfaces given

by λ = constant are ellipsoids, with the λ = 0 surface defining the boundary

of the slice distribution [in fact, α0 in Equation (4.13) is just α (λ = 0) in

Equation (4.14)]. As λ increases these ellipsoids become more spherical, with

λ → R 2 as R → ∞.

The function α (λ), as well as its counterparts in y and z [β (λ) and γ (λ)],

are described in Appendix A, and were implemented in the simulation code

bender to study this problem. Before discussing simulations using this slice

model however, it is instructive to perform an approximate calculation of

the effect of space-charge forces in the last magnet of the chicane. As ar-

gued above, the beam is to a good approximation fully compressed as it

enters the fourth magnet. In configuration space then, the beam consists of

two components, which are being forced together while (approximately) not

moving longitudinally with respect to each other. The space-charge derived

kick (∆px ) on a slice can be found in this case, by integrating the force ap-

199
plied by its counterpart (the slice in the other half of the beam at the same

longitudinal position) as the two slices come together.

For the purpose of this calculation, the x component of the force on one

slice due to its counterpart has the same form as the radial forces found for

uniform density beams in Chapter 2. Namely,




 2mc

2I x
: x<a
γ I0 a 2
2
Fx = . (4.16)



 2mc
2I 1
: x>a
2 γ I0 x

To apply this force, which is plotted in Figure 4.16, we use the value of x

given by the dispersion function in the last magnet,

 2

δp (s − Rbend θbend ) δp
x (s) = [ηx (s)] = , (4.17)
p 2Rbend p

where the equation is written so that s goes from 0 to Rbend θbend in the mag-

net. The momentum kick applied to the slice is then given by

 Rbend θbend
∆px c = Fx [x (s)] ds. (4.18)
0

Using Equations (4.16)-(4.18), for beam parameters relevant to the experiment

at Neptune, gives the 1-2 mrad kick shown in Figure 4.17.

This simple calculation give an estimate for the emittance growth induced

between two beam slices at the same longitudinal position in the bunch. To

make calculations on the full beam, a more sophisticated procedure using

simulations is described below. The model calculation does however, provide

insight into the physical process at work in the last magnet. Note first, that

the model predicts the bifurcation in trace space observed in the experiments.

The space-charge forces between slices in the two beam halves cause them

200
 1.2

0.8

Fmax
0.6
Fx

0.4

0.2
inside outside
ellipsoid ellipsoid
0
0 0.5 1 1.5 2 2.5 3
x
a

Figure 4.16: The force between ellipsoidal slices.

1.5 6

1.0 4 γx
a [mm mrad]

x
[mrad]

0.5 2

0.0 0
0 0.2 0.4 0.6 0.8 1
s
Rbend θbend

Figure 4.17: Model calculation of the angular error and emittance resulting
from space-charge forces in the fourth chicane magnet.

201
to repel, and result, in the model calculation, in a 3-4 mrad difference in the

angles (including the ∆x
on both slices) of the two species.

Also, the extent of this bifurcation depends strongly on the size in x of

the beam slices in the last magnet (which is roughly equal to the size of the

beam at the chicane entrance). As Figure 4.16 shows, slices that overlap in

this model will screen each other, with no transverse force (correctly) applied

when two slices perfectly overlap. This screening is avoided for a longer

period of time, and thus the integrated force is larger, if the beam slices are

made smaller. The increase in emittance with smaller beam size is easily

computed by varying a in the model described by Equations (4.16)-(4.18).

This dependence is shown in Figure 4.18. The normalized emittance in this

calculation is taken at the end of the magnet and is approximately given by

γ∆x
a. Notice that the emittance actually decreases for very small values

of a, where the angular kick cannot be increase further by decreasing beam

size, because the two converging slices do not overlap until very near the end

of the magnet.

Figure 4.16 also shows that, since the derivative of the force outside the

slice boundary is negative, the slices effectively focus each other when the

distance between them is greater than their width. This focusing is nonlinear,

and could lead to slice distributions strongly peaked at the inside edge of

the slice, which would accentuate the bifurcation process further. Thus we

conclude that the estimate of emittance growth deduced from Equation (4.18)

is low, as it does not include this significant nonlinear effect.

Finally, we see that forcing the two beam species together amounts to

202
 6

γx
a [mm mrad]
4

0
0.0 0.5 1.0 1.5 2.0 2.5

a [mm]

Figure 4.18: Model calculation of the emittance resulting from space-charge

forces in the fourth chicane magnet as a function of the horizontal size of
the beam slice.

doing work against the space-charge fields.

  Rbend θbend
dηx δp
W = Fx dx = Fx [x (s)] ds. (4.19)
0 ds p

This energy must, of course, come from the kinetic energy of the beam par-

ticles. Therefore, we expect simulations to show this bifurcation in the mo-

mentum distribution of the compressed beam as well as the transverse phase

space. The bifurcated momentum distribution is in fact seen in parmela

simulations shown below.

To go further with this model, the simulation code named bender (men-

tioned above) was written. The simulated slice distributions are constant

203
density ellipsoids, with boundaries given by the semi-major axes a, b, and c,

as described above. The slices are allowed to change their sizes during the

simulation, and these sizes evolve according to envelope equations similar

to Equation (2.53), with the inclusion of vertical and horizontal edge focus-

ing due to the magnets. The space-charge fields are computed by making a

Lorentz transform to the center of momentum frame of the beam (this is the

same quasi-static approximation made by parmela). In this frame the static

electric fields, at the center of one slice, due to all other slices are computed

via the three dimensional formulas for the ellipsoidal fields given in the ap-

pendix. These fields are then transformed back into the lab frame [104], and

are used to push the slice centroids. The focusing effect described above is

also included in these simulations by finding the average ∆F across the slice

(the nonlinear aspect of this focusing could not be included, as this would

change the form of the slice distribution function, and therefore, make the

field calculations invalid).

The input slice distributions were found, as was done for tredi, by run-

ning parmela up to the entrance of the chicane. The parmela runs in this

case used 105 simulation particles in order to get good statistics on the val-

ues of the slice parameters, such as energy, charge, rms size, rms divergence,

and others.

At this point bender was run for a parameter set matching those of the

Neptune experiment. The primary goal of these runs was to reproduce the

effects observed in the experimental data. The simulated phase space bi-

furcation and dependence of emittance growth on beam size are discussed

below.

204
 0.5

0
x [mm]

-0.5

-1
-1.5 -1 -0.5 0 0.5

δz [mm]

Figure 4.19: bender simulation of the of the beam’s configuration space 50

cm after the chicane exit. The data points define the position of the slice
centroids.

4.2.3 Bifurcations and Size Effects in Simulation

The effects of space-charge in the compressor were seen in results of all

three simulation codes. The most striking results came from bender, as it

employs the heuristic model optimized to illustrate these effects. The first

evidence of space-charge induced beam distortion produced by bender can

be seen in the beam’s configuration space approximately 50 cm downstream

of the chicane (at the location of the slit-mask). This is given for the maxi-

mally compressing case in Figure 4.19, which shows that the configuration

of beam slice centroids is quite different than that predicted by a linear ma-

trix analysis, in which slices should line up along x = 0, as they were at the

205
chicane entrance. Here we see that the head of the beam, which contains

the slices with the highest current, mushrooms out in x as a result of the

space-charge repulsion of these slices.

The bender simulation also shows bifurcation in the trace space of the

maximally compressed beam. This distribution is shown in Figure 4.20a for

the slice centroids. It is also possible to produce a density plot of the trace

space, like those of the experimental data shown in Figure 4.8, since the rms

Twiss parameters of the slices are tracked in the simulation. This plot is

given in part b of Figure 4.20. This bifurcated trace space is quite similar

to those taken from slit data, and is further validation of the space-charge

model used in bender.

Space-charge effects were also seen in the parmela and tredi simula-

tions, although these effects were less pronounced than in bender simula-

tions, and a distinct transverse phase space bifurcation was not seen. This

may be partially due to the low number of simulation particles required to

perform the 3-D space-charge calculations in a reasonable period of time. In-

stead, these effects were displayed most clearly in the beam’s configuration

and longitudinal phase space. Figures 4.21 and 4.22 show a parmela simu-

lation of these spaces after compression. The behavior seen here is similar to

the bender configuration space of Figure 4.19, in that there is a space-charge

induced spreading in the head of the beam, where the density is extremely

high. This time however, we see the spreading occur in particle energy as

well as horizontal position. This results in another bifurcation in the beam;

one in the energy distribution.

206
 1.2

0.8

0.6
x
[mrad]

0.4

0.2

-0.2 (a)

-0.4
0 0.1 0.2 0.3 0.4 0.5

x [mm]

(b)

Figure 4.20: bender simulation of the of the beam’s horizontal phase space
50 cm after the chicane exit. (a) The data points define the position of the
slice centroids. (b) Contour plot of the data from (a). Here the Twiss param-
eters of each slice are used to create their rms ellipses.

207
 22.5

22

21.5

βz γ
21

20.5

20

19.5
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4
δz [mm]

Figure 4.21: parmela simulation of compressed beam’s longitudinal phase

space.

2
x [mm]

-1

-2

-3

-4
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4

δz [mm]

Figure 4.22: parmela simulation of compressed beam’s configuration

space.

208
 700

600

500

400
Counts per Bin

300

200

100

0
19.55 20.3 21.05 21.8
γ

Figure 4.23: parmela simulation of compressed beam’s energy distribution.

This is shown in the double-peaked energy distribution of Figure 4.23,

which is obtained by projecting the phase space distribution of Figure 4.21

onto the γ-axis. As parmela is incapable of including radiative effects, this

energy distribution must be a result of work done against the space-charge

fields as the two parts of the beam are forced together. This effect is also

evident in the longitudinal phase space and energy distribution produced by

tredi simulations, as shown in Figures 4.24 and 4.25, respectively, which

give behavior quite similar to that found with parmela.

As a final comparison of the data with the heuristic space-charge model,

bender was run several times with the PWT phase set for peak compression,

while changing the input transverse size of the slices. The emittance found

209
 22.5

22

21.5

βz γ
21

20.5

20

19.5
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4

δz [mm]

Figure 4.24: tredi simulation of compressed beam’s longitudinal phase

space.

250

200
Counts per Bin

150

100

50

0
19.55 20.05 20.55 21.05 21.55 22.05

Figure 4.25: tredi simulation of compressed beam’s energy distribution.

210
 20

15

εx,n [mm mrad]

10

0
0 0.2 0.4 0.6 0.8 1 1.2

x [mm]

Figure 4.26: bender simulation of the emittance after compression plotted

as a function of input rms size.

at the end of these runs is plotted versus input slice size in Figure 4.26.

This plot shows the same trend predicted by the model calculation and seen

in the experimental data, although it does not “turn on” as quickly as the

model calculation predicts. The experiment seems to be even more sensitive

to this parameter as large emittance growth is observed when the rms size is

reduced from 2 to 1 mm. This sensitivity may be caused by the filamentary

nature of the beam due to emitted beam nonuniformities (from laser or QE

“hot spots”) and a nonideal laser time profile.

211
4.3 Conclusions

The results of the magnetic compression study performed using the Neptune

photoinjector show significant emittance growth and phase space degrada-

tion as the bunch was compressed. To successfully perform this study,

a CTR-based bunch length diagnostic capable of sub-picosecond measure-

ments was implemented. In addition, it was determined that it was neces-

sary to measure and control the (averaged over many shots) PWT phase with

one rf degree precision. Likewise, the required precision in the control and

knowledge of the value of the magnetic fields in the compressor dipoles was

found to be 1%. This precision was required to eliminate residual dispersion

and bend angle dependent steering of the beam exiting the compressor.

Once the method of properly operating the chicane was found, the bunch

length measurements were performed versus compressor setting. The re-

sults of these measurements agreed well with computer simulations, and

the minimum pulse length achieved was 0.6 picoseconds, rms. This value

was also predicted by simulation and is well understood from theory as the

minimum length limited by the curvature imparted to the longitudinal phase

space of the beam by rf fields in the PWT.

At this point the horizontal emittance and trace space of the beam was

measured using the multi-slit mask technique for various magnet and PWT

phase settings. The emittance in these measurements was found to increase

from and uncompressed value of 7 mm mrad (normalized) to a fully com-

pressed value which ranged from 12 to 30 mm mrad, and increased as the size

of the beam decreased. This emittance increase was found to be a function

212
of the final bunch length, and not a strong function of the particular PWT and

magnet settings used to produce this length, as similar results were found by

varying either phase or magnet strength. The beam’s horizontal trace space

was found to bifurcate into two distinct parts, separated in transverse mo-

mentum, as the bunch was compressed. This bifurcation was found to be

linked to the amount of emittance growth observed, and was also a strong

function of transverse spot size.

The experimental conditions were simulated using several different codes,

each designed to include different physical effects. The fact that the results of

the parmela and tredi simulations both agreed well with the experimental

data, and the fact that the elegant (CSR only) simulation showed very little

emittance growth, indicates that the dominant physical emittance growth

mechanism for this experiment was space-charge.

To explore this hypothesis, a heuristic model was developed, in which the

beam is represented by longitudinal slices that interact with each other via

space-charge forces as the beam compresses. This model was used to per-

form a somewhat simplified analytical calculation, and employed in a three

dimensional simulation code. Both the analysis and the simulations using

this model correctly predict unusual trace space bifurcations found in the

experiment, as well as the dependence of emittance growth on transverse

size, supporting the conclusion that space-charge plays a crucial role in this

experiment.

Parmela, tredi, and bender simulations of this experiment all show ev-

idence of the space-charge induced distortions of the beam’s (transverse and

213
longitudinal) phase space, although the results from parmela and tredi are

not as dramatic as those from bender. Part of this inconsistency is undoubt-

edly due to the difficulty in simulating this inherently three dimensional sys-

tem. The consequence of this difficulty is that these simulations must be

performed using a small number of macroparticles, and thus the four di-

mensional (x, px , z, pz ) phase space of the beam is poorly sampled.

While it was argued here that the overall effect of CSR in the system was

small, it is possible that CSR and space-charge effects could act cooperatively

to accentuate the effects observed in this experiment. This possibility seems

to be an interesting avenue of investigation, as recent experimental [105] and

theoretical [106] work shows a CSR instability that can lead to bunching in

the density of the beam both longitudinally and in energy. It may be advan-

tageous in this case to include CSR effects in the bender model, but this is

a topic for future investigation. It should also be noted that the underesti-

mation of beam structure found in this experiment has also been observed

in higher energy experiments. This structure, a filamentation of the beam’s

longitudinal phase space, has been attributed to CSR instability. While the-

oretical and computational models have been developed for this instability,

they apparently underestimate its effect in real experiments. This may, as

in the present experiments, be due to a pre-existing filamentary structure in

the beam’s 6-D phase space before the chicane. This pre-existing structure,

which may be generated at the cathode, is retained in phase space over a large

number of plasma oscillations, and therefore exists through the full length

(only a few oscillations) of an injector. This fact, that the beam does not have

time to reach an equilibrated thermal distribution in the injector, makes the

214
nonlinear analysis of Chapter 2 relevant to the compression process as well.

215
 Appendix A

Photoinjector Simulations

Many different computer simulation codes are employed in this dissertation

to model the beam’s behavior in the photoinjector. These codes are indis-

pensable tools in both the evaluation of theoretical models of the beam and

in understanding the results of experiments. This appendix describes the

most heavily used of these codes in this thesis, and outlines their operating

principles and key features. While the codes homdyn, tredi, and elegant

were used here, these codes were not modified from the versions described

in Refs. [71, 93, 98]. On the other hand, the version of parmela used in

this thesis differs from the widely available version, and the codes norse

and bender were written specifically to investigate the physical processes

discussed in Chapters 2 and 4.

A.1 parmela

The computer program most widely used to simulate photoinjector beams is

parmela (P hase And Radial M otion in E lectron Linear Accelerators). Parmela

was originally developed in 1980 for use on unix machines by Kenneth Cran-

dall at Los Alamos National Laboratory [72]. Later, the code was modified

216
further by Lloyd Young to include the use of photo-cathodes and it contin-

ues to be developed and maintained for the PC platform by the Los Alamos

Accelerator Code Group (LAACG) [107].

Parmela is a versatile program which tracks the beam, represented by a

collection of “macro-particles,” through a user-defined beamline. The beam-

line can be constructed from a wide variety of elements including standing

and traveling wave rf cells, solenoid, dipole, and quadrupole magnets, and

many other element types. It computes the space-charge forces on the beam

via either a two dimensional routine which calculates the forces on a grid,

or a three dimensional point-to-point calculation. The code uses time as the

independent variable, which allows it to correctly model the effects of space-

charge and external forces, especially at the region near the photo-cathode,

where the beam is accelerated from nonrelativistic to relativistic energy very

quickly. In keeping with this concern to accurately model the dynamics of

the beam, parmela also stores the values of the electromagnetic fields on a

grid, and allows the user to input these field values computed by the codes

superfish and poisson (also maintained and distributed by LAACG).

The version of parmela used in this thesis (referred to as “UCLA parmela”)

is based on an early version of the unix code, with a significant number of

additions and modifications implemented by Eric Colby [108]. These addi-

tions include the ability to input three dimensional fields from rf cavities and

solenoid magnets, and an improved routine for generating the initial beam

distribution from the photo-cathode which allows a user-supplied laser dis-

tribution and uses the quite-start [109] algorithm to select the initial phase

space positions of the particles. Perhaps the most useful feature of UCLA

217
parmela is its diagnostic capabilities. The dfiles and fwhm input cards

allow the user to output field and phase space distributions at arbitrary po-

sitions along the beamline, as well as FWHM emittance and beam size calcu-

lations.

A.2 norse

In order to test the analytical predictions of the effects of nonlinear radial

space-charge derived in Chapter 2, the code norse was written. The code

integrates over time the radial space-charge force within a slice, given by
r
2q
Fr (r , z) = 2 2π nb (r̃ , z) r̃ dr̃ . (A.1)
γ r
0

The radial phase space of the slice evolves self-consistently, as Equation (A.1)

is used to push the beam particles at one time step, which in turn changes

the distribution function used to calculate the forces for the next time step.

To avoid numerical noise which would result from binning point particles

in a calculation of the density profile, a distributed macro-particle was used.

The Riccian-type macro-particle used has a finite size and charge distribution

given by
 
qi ri2 + r 2
ρRic (r , ri ) = exp − I0 (r ri ) , (A.2)
2π σi2 2σi2
as introduced in the itaca simulation code [110]. With this type of particle

distribution, the radial distribution of the beam is given simply by summing

the density contributions of each particle at a given point,


ρb (r ) = ρRic (r , ri ) . (A.3)
i

218
Using Equation (A.3) the calculation of the space-charge force of Equation (A.1)

is straightforward.

The use of Riccian particles provides very smooth beam densities, and

therefore it is not feasible to simulate a perfectly hard-edged distribution.

In the context of the study of Chapter 2, that drawback was not very im-

portant since we were interested in investigating distributions were strong

wave-breaking will occur.

A.3 bender

The function of the code bender, and the heuristic model of the beam as a set

of longitudinal slices that fold onto each other in the chicane are described in

Chapter 4. Here, the calculation of the electric fields produced by a uniform

density ellipsoid of charge are given, as these field are used in bender to

derive the space-charge forces between slices as the beam is compressed.

The electrostatic potential both inside and outside the ellipsoid has been

derived by Kellogg [103] using ellipsoidal coordinates. Using Kellogg’s nota-

tion, the basic equation for the ellipsoid of charge is

x2 y 2 z2
+ + = 1, with a2 > b 2 > c 2 , (A.4)
a2 b2 c2

where a, b, and c are the semi-major axes of the ellipsoid. If we define the

functions f (s) and ϕ (s) by

x2 y2 z2
f (s) = + + − 1, (A.5)
a2 + s b2 + s c2 + s
   
ϕ (s) = a2 + s b2 + s c 2 + s , (A.6)

219
then the ellipsoidal coordinates are given by the roots of the cubic equation,

f (s) ϕ (s) = 0. This equation has three real roots (as it must), λ, µ, and ν,

obeying the relation

−a2 ≤ ν ≤ −b2 ≤ µ ≤ −c 2 ≤ λ. (A.7)

As mentioned in Chapter 4, the equation λ = constant defines the surface of

an ellipsoid, with λ = 0 giving the surface of the charge distribution.

The details of the mathematical formalism needed to solve the potential

problem are given by Kellogg and others [101, 111]. The result of these

computation yields the external and internal potential of the ellipsoid,

∞  
x2 y2 z2 ds
Uext = π abcρ0 1− − −  , (A.8)
a2 + s b 2 + s c2 + s ϕ (s)
λ

= −α (λ) x − β (λ) y 2 − γ (λ) z2 + δ (λ) ,

2
(A.9)

and
∞  
x2 y2 z2 ds
Uint = π abcρ0 1− − −  , (A.10)
a2 + s b2 + s c2 + s ϕ (s)
0

= −α0 x − β0 y 2 − γ0 z2 + δ0 ,
2
(A.11)

with
∞ ∞
ds ds
α (λ) = π abcρ0  , and δ (λ) = π abcρ0  ,
(a + s) ϕ (s)
2 ϕ (s)
λ λ
(A.12)

where β (λ) and γ (λ) are found from α (λ) by replacing a with b, and c,

respectively.

The electric fields are, of course, found by differentiating the potential,

and in the case of the internal fields the result is found immediately from

220
Equation (A.11),

Ex,int = 2α0 x,

Ey,int = 2β0 y, (A.13)

Ez,int = 2γ0 z.

The external fields will not in general be a linear function of only one variable,

since Uext is a function of λ. For example, the x-component of the electric

field must be found through

dUext ∂Uext ∂Uext ∂λ

Ex,ext = − =− − . (A.14)
dx ∂x ∂λ ∂x

The remarkable result here is that

 
∞
∂Uext ∂  −f (s)  −f (λ)
= π abcρ0  ds  =  = 0, (A.15)
∂λ ∂λ ϕ (s) ϕ (λ)
λ

since f (λ) = 0 by definition. Therefore, the external fields are given by

Ex,ext = 2α (λ) x,

Ey,ext = 2β (λ) y, (A.16)

Ez,ext = 2γ (λ) z.

From the simulation point of view, one must first compute the ellipsoidal

coordinate λ using the cartesian coordinates and the dimensions of the ellip-

soid, then compute the functions α, β, and γ. The coordinate λ (as well as µ

and ν) is found by solving the cubic equation, w 3 + Aw 2 + Bw + C = 0, with

A = a2 + b 2 + c 2 − x 2 − y 2 − z 2
     
B = c 2 a2 + b 2 − x 2 − y 2 + b 2 a2 − x 2 − z 2 − a2 y 2 + z 2 (A.17)
 
C = c 2 a2 b 2 − b 2 x 2 − a2 y 2 − a2 b 2 z 2

221
In the present case, where a < b < c, the auxiliary quantities Q, R, and θ

given by

A2 − 3B
Q= ,
9
2A3 − 9AB + 27C
R= , (A.18)
 54 
R
θ = cos−1  3 ,
Q

are used to find the ellipsoidal coordinates,

  
θ + 2π A
λ = −2 Q cos − ,
3 3
  
θ − 2π A
µ = −2 Q cos − , (A.19)
3 3
  
θ A
ν = −2 Q cos − .
3 3

All that remains at this point is to evaluate the integrals in Equation (A.12)

to determine α, β, and γ. The form of these functions varies with the sym-

metry of the ellipsoid being considered, with spherical (a = b = c), pro-

late spheroidal (a > b = c), oblate spheroidal (a = b > c), and ellipsoidal

(a > b > c) cases possible. In the case of spherical symmetry the integrals

of Equation (A.12) are easily evaluated, and the expected result, Ex = qx/r 3 ,

is obtained. The results in the other cases were tabulated in Ref. [102], and

are given here.

For the general ellipsoid (a > b > c), α, β, and γ are given by
2
α (λ) = 3/2 [F (φ, k) − E (φ, k)] ,
(a2 − c 2) k2
 
2
2 k
E (φ, k) − k
2 F (φ, k) −  sin φ cos φ ,
β (λ) = 3/2
(a2 − c 2 ) k2 k
2 2 (A.20)
1 − k sin φ
2
  
2 sin φ 1 − k2 sin2 φ
γ (λ) =  − E (φ, k) ,
3/2
(a2 − c 2 ) k
2 cos φ

222
where we use the quantities
  
a −c 
2 2 a2 − b 2 
φ = sin−1  , k= , and k
= 1 − k2 , (A.21)
a2 + λ a2 − c 2

and the elliptic integrals

φ 
E (φ, k) = 1 − k2 sin2 θdθ,
0
(A.22)
φ
dθ
F (φ, k) =  .
1 − k2 sin2 θ
0

For the prolate spheroid (a > b = c), we obtain

2
α (λ) = 3/2 (u − sin φ) ,
(a2 − c 2)
 
2 (a2 − c 2 ) (a2 + λ)
β (λ) = −u , (A.23)
(a2 − c 2 )
3/2
c2 + λ

γ (λ) = β (λ) ,

with 

1 + sin φ
u = ln . (A.24)
1 − sin φ

Finally, for the oblate spheroid (a = b > c), we have

2
α (λ) = 3/2 (φ − sin φ cos φ) ,
(a2 − c 2 )

β (λ) = α (λ) , (A.25)

2
γ (λ) = 3/2 (tan φ − φ) .
(a2 − c 2 )

223
Figure A.1: One component of the electric field (Ex ) produced by a uniformly
charged ellipsoid, and plotted for points in the x − y plane. Inside the distri-
bution the field is linear in x and independent of y and z. For this example
the semi-major axes of ellipsoid are (a, b, c) = (5, 1, 2)

224
 References

[1] Keith R. Symon. Mechanics (Addison Wesley Longman, Inc., 1971).

[2] J. D. Lawson. The Physics of Charged-particle Beams (Clarendon Press,

Oxford, 1977).

[3] Martin Reiser. Theory and Design of Charged Particle Beams (Wiley,
New York, 1994).

[4] James B. Rosenzweig. Fudamentals of Beam Physics (Oxford University

Press, Oxford, 2002).

[5] Claude Lejeune and Jean Aubert. “Emittance and brightness: Defi-
nitions and measurements.” In Advances in Electronics and Electron
Physics, Supplement 13A, p. 159 (1980).

[6] P. M. Lapostolle. “Possible emittance increase through filamentation

due to space charge in continuous beams.” IEEE Trans. Nucl. Sci., 18,
p. 1101 (1971).

[7] F. J. Sacherer. “Rms envelope equations with space charge.” IEEE Trans.
Nucl. Sci., 18, p. 1105 (1971).

[8] M. G. Van der Wiel. “Applications of high-brightness electron beams.”

In Proceedings of the ICFA Advanced Accelerator Workshop on the
Physics of High Brightness Beams, Los Angeles, 1999, edited by J. B.
Rosenzweig and L. Serafini, p. 3 (World Scientific, Singapore, 2000).

[9] A. M. Kondratenko and E. L. Saldin. “Generation of coherent radiation

by a relativistic electron beam in an ondulator.” Part. Accel., 10, p. 207
(1980).

[10] R. Bonifacio, C. Pellegrini, and L. M. Narducci. “Collective instabilities

and high-gain regime in a free electron laser.” Opt. Commun., 50, p.
373 (1984).

[11] C. Pellegrini. “Fel collective instability.” UCLA Lecture Notes (1998).

[12] The LCLS Design Study Group. “Linac coherent light source (lcls) de-
sign study report.” Technical Report SLAC Report No. SLAC-R-0521,
Stanford Linear Accelerator Center (1998).

225
[13] Sven Reiche. Private communication.

[14] C. A. Brau and W. B. Mori. “Working group 3 summary: Exotic source

physics: photon-assited field emission, spin polarization, cathode-less
(plasma) sources, etc.” In Proceedings of the ICFA Advanced Acceler-
ator Workshop on the Physics of High Brightness Beams, Los Angeles,
1999, edited by J. B. Rosenzweig and L. Serafini, p. 393 (World Scien-
tific, Singapore, 2000).

[15] J. R. Pierce. Theory and Design of Electron Beams (Van Nostrand, 1954).

[16] M. Reiser, S. Bernal, A. Dragt, M. Venturini, J. G. Wang, et al. “Design

features of a small electron ring for study of recirculating space-charge
dominated beams.” Fusion Eng. Des, 32-33, p. 293 (1996).

[17] D. B. Langmuir. Proc. Inst. radio Engrs, 25, p. 977 (1937).

[18] O. W. Richardson. Phil. Mag., 28, p. 633 (1914).

[19] S. Dushman. “Thermionic emission.” Rev. Modern Phys., 2, p. 381

(1930).

[20] Patrick G. O’Shea. “Rf photoinjectors.” In Proceedings of the ICFA Ad-

vanced Accelerator Workshop on the Physics of High Brightness Beams,
Los Angeles, 1999, edited by J. B. Rosenzweig and L. Serafini, p. 17
(World Scientific, Singapore, 2000).

[21] D. T. Palmer. The next generation photoinjector. Ph.D. thesis, Stanford

University (1998).

[22] M. Borland. Technical Report SLAC-Report-402, Stanford Linear Accel-

erator Center (1991).

[23] K. J. Kim. “Rf and space-charge effects in laser-driven rf electron guns.”

Nucl. Instrum. Methods Phys. Res., Sect. A, 275, p. 201 (1989).

[24] J. E. Clendenin and G. A. Mulhollan. In Quantum Aspects of Beam

Physics, edited by P. Chen, p. 254 (World Scientific, Singapore, 1999).

[25] D. T. Palmer, X. J. Wang, R. H. Miller, M. Babzien, I. Ben-Zvi, et al. “Emit-

tance studies of the bnl/slac/ucla 1.6 cell photocathode rf gun.” In
Proceedings of the 1997 Particle Accelerator Conference, p. 2687 (IEEE,
Piscataway, NJ, 1997).

226
[26] J. E. Clendenin, T. Kotseroglou, G. A. Mulhollan, D. T. Palmer, and J. F.
Schmerge. “Reduction of thermal emittance of rf guns.” Technical
Report SLAC-PUB-8284, Stanford Linear Accelerator Center (1999).

[27] M. Ferrario, J. E. Clendenin, D. T. Palmer, J. B. Rosenzweig, and L. Ser-

afini. “Homdyn study for the lcls rf photo-injector.” In Proceedings of
the ICFA Advanced Accelerator Workshop on the Physics of High Bright-
ness Beams, Los Angeles, 1999, edited by J. B. Rosenzweig and L. Ser-
afini, p. 534 (World Scientific, Singapore, 2000).

[28] J. S. Fraser and R. L. Sheffield. “High-brightness injectors of rf-driven

free-electron lasers.” IEEE J. Quantum Electron., 23, p. 1489 (1987).

[29] S. C. Hartman and J. B. Rosenzweig. “Ponderomotive focussing in ax-

isymmetric rf linacs.” Phys. Rev. E, 47, p. 2031 (1993).

[30] J. B. Rosenzweig and L. Sarafini. “Transverse particle motion in radio-

frequency linear accelerators.” Phys. Rev. E, 49, p. 1599 (1993).

[31] H. G. Kirk, R. Miller, and D. Yeremian. “Electron guns and preinjectors.”

In Handbook of accelerator physics and engineering, p. 99 (World Sci-
entific, 1999).

[32] C. E. Clayton, C. Joshi, K. A. Marsh, C. Pellegrini, and J. B. Rosenzweig.

“The neptune facility for 2nd generation advanced accelerator experi-
ments.” In Proceedings of the 1997 Particle Accelerator Conference, p.
678 (IEEE, Piscataway, NJ, 1997).

[33] C. E. Clayton, C. Joshi, K. A. Marsh, C. Pellegrini, and J. B. Rosen-

zweig. “Second generation beatwave experiments at ucla.” Nucl. In-
strum. Methods Phys. Res., Sect. A, 410, p. 378 (1998).

[34] J. B. Rosenzweig, S. G. Anderson, K. Bishofberger, X. Ding, A. Murokh,

et al. “The neptune photoinjector.” Nucl. Instrum. Methods Phys. Res.,
Sect. A, 410, p. 437 (1998).

[35] C. E. Clayton, C. Joshi, C. Darrow, and D. Umstadter. “Relativistic

plasma-wave excitation by collinear optical mixing.” Phys. Rev. Lett.,
54, p. 2343 (1985).

[36] N. Barov, J. B. Rosenzweig, M. E. Conde, W. Gai, and J. G. Power. “Ob-

servation of plasma wakefield acceleration in the underdense regime.”
Phys. Rev. ST Accel. Beams, 3, p. 011301 (2000).

227
[37] P. Musumeci, C. Pellegrini, J. B. Rosenzweig, A. Varfolomeev, S. Tol-
machev, et al. “On the ifel experiment at the ucla neptune lab.” In
Proceedings of the 2001 Particle Accelerator Conference, p. 4008 (IEEE,
Piscataway, NJ, 2001).

[38] N. Barov, M. E. Conde, W. Gai, and J. B. Rosenzweig. “Propagation of

short electron pulses in a plasma channel.” Phys. Rev. Lett., 80, p. 81
(1998).

[39] J. B. Rosenzweig, N. Barov, and E. Colby. “Pulse compression in rf

photoinjectors: Applications to advanced accelerators.” IEEE Trans.
Plasma Sci., 24, p. 409 (1996).

[40] D. T. Palmer, X. J. Wang, I. Ben-Zvi, and R. H. Miller. “Beam dy-

namics enhancement due to accelerating field symmetrization in the
bnl/slac/ucla 1.6 cell s-band photocathode rf gun.” In Proceedings of
the 1997 Particle Accelerator Conference, p. 2846 (IEEE, Piscataway, NJ,
1997).

[41] S. H. Kong, J. Kinross-Wright, D. C. Nguyen, and R. L. Sheffield. “Cesium

telluride photocathodes.” J. Appl. Phys., 77, p. 6031 (1995).

[42] D. T. Palmer, S. G. Anderson, and J. B. Rosenzweig. “Single crystal

cooper photo-cathode in the bnl/slac/ucla 1.6 cell rf gun.” In Proceed-
ings of the ICFA Advanced Accelerator Workshop on the Physics of High
Brightness Beams, Los Angeles, 1999, edited by J. B. Rosenzweig and
L. Serafini, p. 439 (World Scientific, Singapore, 2000).

[43] R. Zhang, P. Davis, G. Hairapetian, M. Hogan, C. Joshi, et al. “Initial

operation of the ucla plane wave transformer (pwt linac).” In 1995
Particle Accelerator Conference, p. 1102 (IEEE, 1995).

[44] L. C. Maier and J. C. Slater. J. Appl. Phys., 23, p. 68 (1952).

[45] D. T. Palmer, R. H. Miller, H. Winick, X. J. Wang, K. Batchelor, et al.

“Microwave measurements of the bnl/slac/ucla 1.6 cell photocathode
rf gun.” In Proceedings of the 1995 Particle Accelerator Conference, p.
982 (IEEE, Piscataway, NJ, 1995).

[46] Anthony E. Siegman. Lasers (University Science Books, Sausalito, CA,

1986).

228
[47] Pietro Musumeci. Private communication.

[48] X. J. Wang, M. Babszien, I. Ben-Zvi, X. Y. Chang S. Pjerov, and M. Woodle.

“High-rep rate photocathode injector for lcls.” In Proceedings of the
2001 Particle Accelerator Conference, p. 2233 (IEEE, Piscataway, NJ,
2001).

[49] X. J. Wang. “Progress and future directions in brightness electron beam

sources.” In Proceedings of the 2001 Particle Accelerator Conference,
p. 81 (IEEE, Piscataway, NJ, 2001).

[50] M. Hernandez, A. Fisher, D. Meyerhofer, R. Miller, D. T. Palmer, et al.

“Emittance measurements for the slac gun test facility.” In Proceedings
of the 1997 Particle Accelerator Conference, p. 2840 (IEEE, Piscataway,
NJ, 1997).

[51] George Caryotakis. “The klystron: A microwae source of surprising

range and endurance.” Technical Report SLAC-PUB-7731, Stanford Lin-
ear Accelerator Center (1998).

[52] S. G. Anderson, J. B. Rosenzweig, G. P. LeSage, and J. K. Crane. “Space-

charge effects in high brightness electron beam emittance measure-
ments.” Phys. Rev. ST Accel. Beams, 5, p. 014201 (2002).

[53] U. Happek, A. J. Sievers, and E. B. Blum. “Observation of coherent tran-

sition radiation.” Phys. Rev. Lett., 67, p. 2962 (1991).

[54] A. Murokh, J. B. Rosenzweig, M. Hogan, H. Suk, G. Travish, et al. “Bunch

length measurement of picosecond electron beam from a photoinjector
using coherent transition radiation.” Nucl. Instrum. Methods Phys. Res.,
Sect. A, 410, p. 452 (1998).

[55] Bergoz Precision Beam Instrumentation. See http://www.bergoz.

com/.

[56] G. Travish. Experimental Requirements of a Self-Amplified Spontaneous

Emission Test System: Design, Construction, Simulation and Analysis of
the UCLA High Gain Free Electron Laser. Ph.D. thesis, University of
California, Los Angeles (1996).

[57] W. S. Graves, E. D. Johnson, and P. G. O’Shea. “A high resolution electron

beam profile monitor.” In Proceedings of the 1997 Particle Accelerator
Conference, p. 1993 (IEEE, Piscataway, NJ, 1997).

229
[58] National Instruments. LabVIEW User Manual. Part No. 320999D-01
(2001).

[59] See http://www.transera.com/htbasic/tutgpib.html.

[60] See http://www.kscorp.com/www/products/frame.htm.

[61] B. E. Carlsten. “New photoelectric injector design for the los alamos
national laboratory xuv fel accelerator.” Nucl. Instrum. Methods Phys.
Res., Sect. A, 285, p. 313 (1989).

[62] Luca Serafini and J. B. Rosenzweig. “Envolope analysis of intense rela-

tivistic quasilaminar beams in rf photoinjectors: A theory of emittance
compensation.” Phys. Rev. E, 55, p. 7565 (1997).

[63] X. Qiu, K. Batchelor, I. Ben-Zvi, and X. J. Wang. “Demonstration of emit-

tance compensation through the measurement of the slice emittance
of a 10-ps electron bunch.” Phys. Rev. Lett., 76, p. 3723 (1996).

[64] T. P. Wangler. RF Linear Accelerators (Wiley, New York, 1998).

[65] L. Brillouin. Phys. Rev., 67, p. 260 (1945).

[66] B. E. Carlsten. Phys. Rev. E, 60, p. 2280 (1999).

[67] O. A. Anderson. “Internal dynamics and emittance growth in space-

charge-dominated beams.” Part. Accel., 21, p. 197 (1987).

[68] I. Hoffman and J. Struckmeier. Part. Accel., 21, p. 69 (1987).

[69] T. P. Wangler, K. R. Crandall, R. S. Mills, and M. Reiser. IEEE Trans. Nucl.

Sci., 32, p. 2196 (1985).

[70] Patrick G. O’Shea. Phys. Rev. E, 57, p. 1081 (1998).

[71] M. Ferrario. “Multi-bunch energy spread induced by beam loading in a

standing wave structure.” Part. Accel., 52, p. 1 (1996).

[72] L. Young and J. Billen. “Parmela.” Technical Report LA-UR-96-1835,

Los Alamos National Laboratory (1996).

[73] L. Serafini and J. B. Rosenzweig. “Optimum operation of split rf pho-

toinjectors.” In Proceedings of the 1997 Particle Accelerator Confer-
ence, p. 2876 (IEEE, Piscataway, NJ, 1997).

230
[74] B. E. Carlsten, John C. Goldstein, Patrick G. O’Shea, and Eric J. Pitcher.
Nucl. Instrum. Methods Phys. Res., Sect. A, 331, p. 791 (1993).

[75] G. P. LeSage, S. G. Anderson, T. E. Cowan, J. K. Crane, T. Ditmire, et al. In

Advanced Accelerator Concepts: Ninth Workshop, edited by P. L. Cole-
stock and S. Kelley, AIP Conf. Proc. No. 569 (AIP, New York, 2001).

[76] R. Q. Twiss and N. H. Frank. Rev. Sci. Instrum., 20, p. 1 (1949).

[77] S. C. Hartman, N. Barov, C. Pellegrini, S. Park, J. B. Rosenzweig, et al.

“Initial measurements of the ucla photoinjector.” Nucl. Instrum. Meth-
ods Phys. Res., Sect. A, 340, p. 219 (1994).

[78] Helmut Wiedemann. Particle Accelerator Physics: Basic Principles and

Linear Beam Dynamics (Springer-Verlag, New York, 1993).

[79] G. R. Hays. In Proceedings of the Conference on Lasers and Electro-

Optics, San Fransico, p. 293 (Optical Society of America, Washington
DC, 2000).

[80] S. G. Anderson and J. B. Rosenzweig. “Non-equilibrium transverse mo-

tion and emittance growth in space-charge dominated beams.” Phys.
Rev. ST Accel. Beams, 3, p. 094201 (2000).

[81] L. Serafini. IEEE Trans. on Plasma Science, 24, p. 421 (1996).

[82] M. James, J. Clendenin, S. Ecklund, R. Miller, J. Sheppard, et al. “Up-

date on the high-current injector for the stanford linear collider.” IEEE
Trans. Nucl. Sci., NS-30, p. 2992 (1983).

[83] Ya. S. Derbenev, J. Rossbach, E. L. Saldin, and V. D. Shiltsev. “Mi-

crobunch radiative tail-head interaction.” Technical Report TESLA-FEL-
95-05, DESY (1995).

[84] E. L. Saldin, E. A. Schneidmiller, and M. Yurkov. “Klystron instability of

a relativistic electron beam in a bunch compressor.” Technical Report
TESLA-FEL-2002-02, DESY (2002).

[85] H. Braun, R. Corsini, L. Groening, F. Zhou, A. Kabel, et al. “Emittance

growth and energy loss due to coherent synchrotron radiation in a
bunch compressor.” Phys. Rev. ST Accel. Beams, 3, p. 124402 (2000).

231
[86] H. Braun, R. Corsini, L. Groening, F. Zhou, A. Kabel, et al. “Emittance
growth and energy loss due to coherent synchrotron radiation in the
bunch compressor of the clic test facility (ctf ii).” In Proceedings of
EPAC 2000, p. 471 (2000).

[87] M. Dohlus, A. Kabel, and T. Limberg. “Optimal beam optics in the ttf-
fel bunch compression sections: Minimizing the emittance growth.” In
Proceedings of the 1999 Particle Accelerator Conference, p. 1650 (IEEE,
1999).

[88] M. Borland. “Aps csr experiments and comparison with simula-

tion.” ICFA Beam Dynamics mini workshop: Coherent Synchrotron
and its impact on the beam dynamics of high brightness electron
beams. See http://www.desy.de/csr/csr_workshop_2002/csr_
workshop_2002_index.html (2002).

[89] Integrated Engineering Software. See http://www.integratedsoft.

com/.

[90] H. c. Lihn, P. Kung, C. Settakorn, and H. Weidemann. “Measurement of

subpicosecond electron pulses.” Phys. Rev. E, 53, p. 6413 (1996).

[91] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery. Numer-

ical Recipes in FORTRAN (Cambridge University Press, 1986), second
edition.

[92] A. Tremaine, J. B. Rosenzweig, S. G. Anderson, P. Frigola, M. Hogan, et al.

“Observation of self-amplified spontaneous emission-induced electron
beam microbunching using coherent transition radiation.” Phys. Rev.
Lett., 81, p. 5816 (1998).

[93] F. Ciocci, L. Giannessi, A. Marranca, L. Mezi, and M. Quattromini. “Self-

consistent three-dimensional rf-gun dynamics integration based on the
lienard-wiechert retarded potentials.” Nucl. Instrum. Methods Phys.
Res., Sect. A, 393, p. 434 (1997).

[94] Richard Talman. “Novel relativistic effect important in accelerators.”

Phys. Rev. Lett., 56, p. 1429 (1986).

[95] B. E. Carlsten and T. O. Raubenheimer. “Emittance growth of bunched

beams in bends.” Phys. Rev. E, 51, p. 1453 (1995).

232
 [96] E. P. Lee. Part. Accel., 25, p. 241 (1990).

[97] R. Li. “Progress on the study of csr effects.” In Proceedings of the

ICFA Advanced Accelerator Workshop on the Physics of High Brightness
Beams, Los Angeles, 1999, edited by J. B. Rosenzweig and L. Serafini, p.
369 (World Scientific, Singapore, 2000).

[98] M. Borland. “User’s manual for elegant.” See http://www.aps.anl.

gov/asd/oag/manuals/elegant_ver14.5/elegant.html (2001).

[99] I. M. Kapchinsky and V. V. Vladimirski. In Proc. International Confer-

ence on High Energy Accelerators, p. 274 (CERN, Geneva, 1959).

[100] P. M. Lapostolle. “Effets de la charge d’espace dans un accelerateur a

protons.” Technical Report AR/Int. SG/65-15, CERN (1965).

[101] W. E. Byerly. An Elementary Treatise on Fourier’s Series and Spherical,

Cylindrical and Ellipsoidal Harmonics with Applications to Problems in
Mathematical Physics (Ginn & Company, Boston, 1893).

[102] L. B. Tuckerman. “Inertia factors of ellipsoids for use in airship design.”

Technical Report 210, National Advisory Committee for Aeronautics
(1926).

[103] O. D. Kellogg. Foundations of Potential Theory (F. Ungar Pub. Co., New
York, 1929).

[104] J. D. Jackson. Classical Electrodynamics (John Wiley & Sons, New York,
1962).

[105] W. Graves. “Microbunching and csr experiments at bnls source de-

velopment lab.” ICFA Beam Dynamics mini workshop: Coherent Syn-
chrotron and its impact on the beam dynamics of high brightness elec-
tron beams. See http://www.desy.de/csr/csr_workshop_2002/
csr_workshop_2002_index.html (2002).

[106] G. Stupakov and S. Heifets. “Beam instability and microbunching due

to coherent synchrotron radiation.” Phys. Rev. ST Accel. Beams, 5, p.
054402 (2002).

[107] Los Alamos Accelerator Code Group. See http://laacg1.lanl.gov/.

233
[108] Eric Colby. “Modifications to parmela, revisited.” Technical report,
UCLA Particle Beam Physics Lab. See http://www.stout.physics.
ucla.edu/docs/parmela.pdf (1998).

[109] S. Reiche. Numerical Studies for a Single Pass High Gain Free-Electron
Laser. Ph.D. thesis, Hamburg University (1999).

[110] Luca Serafini. Private communication.

[111] S. Chandrasekhar. Ellipsoidal Figures of Equilibrium (Yale University

Press, New Haven, 1969).

234