Accel Notes
Accel Notes
First-
and
Second-Order
Matrix
Theory
for the
Design Charged
of Beam Particle
Transport
Systems
and
Spectrometers
Karl
L.
Brown
SLAC
Report-75 1982
June
Under
contract of
with Energy
the
Department Contract
DE-AC03-76SFO0515
STANFORD Stanford
L INEAR
ACCELERATOR . Stanford
CENTER California
University
This document and the material and data contained therein, was develUnited States Government. Neither the oped under sponsorship of the the Leland Stanford the Department of Energy, nor United States nor nor their respective contracJunior University, nor their employees, tors subcontractors, or their employees, makes any warranty, express responsibility for accuracy, or implied, or assumes any liability or usefulness of any information, apparatus, product or completeness or that its use will not infringe priprocess disclosed, or represents any product, rights. its manufacturer, or vately-owned Mention of approval, disapintended to, imply shall not, nor is it suppliers A royalty-free, nonexfitness for use. any particular proval, or clusive right to use and disseminate same for any purpose whatsoever, the University. to the United States and expressly reserved is 7/82
I
1
I
A First- and Second-Order Matrix Theory for the Design of Beam Transport Systems and Charged Particle Spectrometers*+
KARL L. BROWN
Contents
1. Introduction . 11. A General First- and Second-Order Theory of Beam Transport Optics 1. The Vector Differential Equation of Motion . 2. The Coordinate System . . . 3. Expanded Form ofa Magnetic Field Having Median Plane Symmetry 4. Field Expansion to Second Order Only . . oft~ and ~ with Pure Quadruple and Sextupole Fields 5. Identification 6. The Equations of Motion in Their Final Form to Second Order 7. The Description of the Trajectories and the Coefficients of the Taylors Expansion . . . . . 8. Transformation from Curvilinear Coordinates to a Rectangular Coordinate System and TRANSPORT Notation . 9. First- and SecQnd-Order Ma(rix Formalism of Beam Transport Optics IIl. Reduction of the General First- and Second-Order Theory to the Case of theldeal Magnet. 1. Matrix Elements for a Pure Quadruple Field . 2. Matrix Elements for a Pure Sextupole Field . 3. First- and Second-Order Matrix Elements for a Curved, Inclined Magnetic Field Boundary . . 4. Matrix Elements for a Drift Distance IV. Some Useful First-Order Optical Results Derived from the General Theory of Section 11 . 1. First-Order Dispersion 2. First-Order Path Length
* Work supported
by the QepaFtment of Energy, contract DE-AC03-76SFO0515. in Particle
72 78 82 82 85 89 91 91 93 98 102
107
114
115
16 18 19
20
I 20
t Permission to reprint this article, published in Advances granted by John Wiley and Sons, Inc. Physics Journal,
72
K. L. BROWN 121 121 121 122 122 1~~ 123 123 124
A. Achromaticity . B. Isochronicity 3. First-Order Imaging 4. Magnification 5. First-Order Momentum Resolution 6. Zero Dispersion 7. Focal Length. . 8. Evaluation of the First-Order Matrix for Ideal Magnets 9, The R Matrix Transformed to the Principal Planes V. Some General Second-Order Theorems Derived from the General Theory of Section 11 . 1. Optical Symmetries in )Z = + Magnetic Systems VI, An Approximate Evaluation of the Second-Order Aberrations for HighEnergy Physics l. Case I . 2. Case II. . 3. Case III . References . . . . . . . .
I. Introduction
Since the invention of the alternating gradient principle and the subsequent design of the Brookhaven and CERN proton-synchrotrons
based on this principle, there has been a rapid matical and physical techniques applicable to In this report a matrix algebra formalism will essential principles governing the design of transport systems, with a particular emphasis evolution of the mathecharged particle optics. be used to develop the charged particle beam on the design of high-
energy magnetic spectrometers. A notation introduced by John Streib(l) has been found to be ~seful in conveying the essential physical principles dictating the design of such beam transport systems. ln particular to first order, the momentum dispersion, the momentum resolution, the particle path length, and the necessary and sufficient conditions for zero dispersion, achromaticity, and isochronicity may all be expressed as simple integrals of particular first-order trajectories (matrix elements) characterizing a system. This formulation provides direct physical insight into the design of beam transport systems and charged particle spectrometers. An intuitive grasp of the mechanism of second-order aberrations also results from this formalism; for example, the effects of magnetic symmetry on the minimization or elimination ofsecond-order aberrations is immediately apparent.
SYSTEM
AND
SPECTROMETER
DESIGN
73
The equations of motion will be derived and then the matrix formalism tintroduced, developed, and evolved into useful theorems. Physical examples will be given to illustrate the applicability of the
formalism to the design of specific spectrometers. It is hoped that the information supplied will provide the reader with the necessary tools so that he can design any beam transport system or spectrometer suited to his particular needs. The theory has been developed to second order in a Taylor expan-
sion about a central trajectory, characterizing the system. This seems to be adequate for most high-energy physics applications. For studying details beyond second order, we have found computer ray tracing programs to be the best technique for verification of matrix calculations, and as a means for further refinement of the optics if needed. .ln the design of actual systems for high-energy beam transport applications, it has proved convenient to express the results via a multipolcexpansion about a central trajectory. in this expansion, the constant tcrrn proportional to the Iicld strc]lgth at the ccntra] trajectory is the dipole term. The term proporti(~n~~l to the first derivative of the field (with respect to the transverse dinlcnsi(lns) about the central tr:ijcctory is a quadrupolc term and the second derivative with respect to the transverse dimensions is a sextupole tcrrn, etc. A considerable design simplification results at high cncrgics if the dipole, quadruple, and sextupole functions arc physically separated stich that cross-product terms among them do not appear, and if the fringing field effects are small compared to the contributions of the multipole elements comprising the systcm. At the risk of oversimplification, the basic function of the multipole elements may bc identified in the following way: The purpose of the dipole element(s) is to bend the central trajectory of the systcm and disperse the beam; th;it is, it is the means of providing the first-order nlomcntLlnl dispersion for the systcm. The quadruple etement(s) generate the first-order irn:lging. The sextupole terms couple with the second-order aberrations; and a scxtl]polc element introduced into the systcm is :1 mechanism for minimizing or etirninating a particuliir second-order aberration that may h:lve been
generated by dipole or quadrupotc elements. Quadruple elements may bc introduced in any CJIIC of (hrcc chartictcristic fOrms: (/) via
an actu:ll physicul qu:tdrupolc consisting (>f four poles such that a first fitld derivative exists in the fictd expansion about the central traicct(~ry~ (2)-via :i rotated input or output face of a
74
K. L. BROWN
bending magnet; and (3) via a transverse field gradient in the dipole elements of the system. Clearly any one of these three fundamental mechanisms may be used as a means of achieving first-order imaging in a system. Of course dipole elements will tend to image in the radial bending plane independent of whether a transverse field derivative does or does not exist in the system, but imaging perpendicular to the plane of bend is not possible without the introduction of a first-field derivative. In addition to their fundamental purpose, dipoles and quadruples will also introduce higher-order aberrations. If these aberrations are second order, they may be eliminated or at least modified by the introduction of sextupole elements at appropriate locations. In regions of zero dispersion, a sextupole will couple with and modify only geometric aberrations. However, in a region where momentum dispersion is present, sextupoles will also couple with and modify chromatic aberrations. Similar to the quadruple, a sextupole element may be generated in one of several ways, first by incorporating an actual sextupole, that is, a six-pole magnet, into the system. However, any mechanism which introduces a second derivative of the field with respect to the transverse dimensions is, in effect, introducing a sextupole component. Thus a second-order curved surface on the entrance or exit face of a bending magnet or a second-order transverse curvature on the pole surfiaces of a bending magnet is also a sextupole component. As illustrations of systems possessing dipole, quadruple, and sextupole elements, consider _the n = + double-focusing spectrometer which is widely used for low- and medium-energy physics applications. Clearly there is a dipole element resulting from the presence of a magnetic field component along the central trajectory of the spectrometer. A distributed quadruple element exists as a consequence of the ~~= + field gradient. In this particular case, since the transverse imaging forces are proportional to nl/2 and the radial imaging forces are proportional to (1 n)l2, the restoring forces are equal in both planes, hence the reason for the double focusing properties. In addition to the first derivative of the field )7 = (rO/BO)(6B/6r), there are usually second- and higher-order transverse field derivatives present. The second derivative of the field ~ = +(r~/BO)(62B/br2j introduces a distributed sextupole along the entire length of the spectrometer. Thus to second order a typical n = + spectrometer consists of a single dipole with a distributed,
SYSTEM
AND
SPECTRO
METER
DESIGN
75
quadrupoie andsextupole superimposed along the entire length of the dipole element. Higher-order multipoles mayalso represent, but will be ignored in this discussion. [n the preceding example the dipole, quadruple, and sextupole functions are integrated in the same magnet. However, in many highenergy applications itis often more economical to use separate magnetic elements for each of the multipole functions. Consider also the SLAC
spectrometers which provide examples of solutions which combine the multipole functions into a single magnet as well as solutions using separate multipole elements. Three spectrometers have been designed: one for a maximum energy of 1.6 GeV/c to study large backward angle scattering processes, a second for 8 GeV/c to study intermediate forward angle production processes, and finally a 20-GeV/c spectrometer for small forward angle production. All of these instruments are to be used in conjunction with primary electron and gamma-ray energies in the range of 10-20 GeV/c. The 1.6-GeV/c instrument (Fig. 1) is a single magnet, bending the
Focal
plane
Po ~
&
Uniform field regions
Po
\_
Shadedareas indicate
regions
possessing sextupole components
of the field
\
$ {
4L
PI
-. FIG. 1. 1.6-GeVlc spectrometer.
76
K. L. BROWN
central trajectory a total of 90, thus constituting the dipole contribution to the optics of the system. Two quadruple elements are present in the magnet; i.e.. input and output pole faces of the magnet are rotated so as to provide transverse focusing. and the 90 bend provides radial focusing via the (1 }Z)12factor characteristic of any dipole magnet. The net optical result is point-to-point imaging in the plane of bend and parallel-to-point imaging in the plane transverse to the plane of bend. The solid angle and resolution requirements of the 1.6-GeV/c spectrometer are such that three sextupole components are needed to achieve the required performance. In this application, the sextupoles are generated by machining an appropriate transverse second-order curvature on the magnet pole face at three different locations along the 90 bend of the system. In summary, the 1.6-GeV/c spectrometer consists of one dipole, bending a total of 902, two quadruple elements, and a sextupole triplet with the quadruple and sextupole strengths chosen to provide the first- and second-order properties demanded of the system.
. \
\ >~
>, ?-.\
I +~Total
, :; ,. , o -gg----o ::4:
,,
Q3 ~~
15
path -
FIG. 2.
Magnet arrangement,
8-GeV/c
spectrometer.
Optically, the 8-GeV/c spectrometer (Fig. 2) is relatively simple. It consists of two dipoles, each bending 159, making a total of a 30 bend, and three quadruples (two preceding and one following the dipole elements) to provide point-to-point imaging in the plane of bend and parallel-to-point imagitig ii the plane transverse to the bending plane. The solid angle and resolution requirements of the instrument are
SYSTEM
AND
SPECTROMETER
DESIGN
77
sufficiently modest that no sextupole components are needed. The penalty paid for not adding sextupole components is that the focal plane angle with respect to the optic axis at the end of the system is a relatively small angle (13.70). With the addition of one sextupole element near the end of the system, the focal plane could have been rotated to a much larger angle. However, the 13.7 angle was acceptable for the focal plane counter array and as such it was ultimately decided to omit the additional sextupole element. The 20-GeV/c spectrometer (Fig. 3) is a more complex design. The an d B. d] twice that of the 8-GeVc f spectrometer. The final instrument is composed of four dipole elements (bending magnets), two bending in one sense and the other two bending in the opposite sense, so the beam emanating from the instrument is parallel to the incident primary particles. The first-order imaging is achieved via four quadruples. The chromatic aberrations generated by the quadruples in this system are more serious than in the 8-GeV/c case because of an intermediate image required at the midpoint of the system. As a result, the focal plane angle with respect to the central trajectory would have been in the range of24. As a consequence, sextupoles were introduced in order to rotate the focal plane to a more satisfactory angle for the counter array. A final compromise placed the focal plane angle at 45 with respect to the optic axis of the system via the introduction of three sextupoles. Thus the 20-GeV/c spectrometer consists of four dipoles, with an intermediate crossover following the first two dipoles. a quadruple triplet to achieve first-order imaging, and a sextupole triplet to compensate for the chromatic aberrations introduced by the quadrupoles. Optically, the 20-GeV/c spectrometer is very similar to the increased momentum requires
Focal plane
Q Q Q Q
B
FIG. 3. 20-GeV/c
spectronleter.
78
K. L. BROWN
1.6-GeV/c spectrometer and yet physically it is radically different because of the method of introducin& the various multipole components. Having provided some representative examples of spectrometer design, we now wish to introduce and develop the theoretical tools for creating other designs.
Theory of
The fundamental objective is to study the trajectories described b} charged particles in a static magnetic field. To maintain the desired generality, only one major restriction will be imposed on the field configuration: Relative to a plane that will be designated as the magnetic midplane, the magnetic scalar potential v shall be an odd function in the transverse coordinate ~ (the direction perpendicular to the midplane), i.e., P(.Y,}, ?) = v(.Y, j, t). This restriction greatly simplifies the calculations, and from experience in designing beam transport systems it appears that for most applications there is little, if any, advantage to be gained from a more complicated field pattern. The trajectories ~~ill be described by means of a Taylors expansion about a particular trajectory (which lies entirely within the magnetic midplane) designated henceforth as the central trajectory. Referring to Figure 4, the coordinate I is the arc length _measured along the central trajectory; and ,~> ~Iand 1 form a right-handed curvilinear coordinate system. The results will be valid for describing trajectories lying close to and making small angles with the central trajectory. The basic steps in formulating the solution to the problem are as follows : 1. A general vector differential equation is derived describing the trajectory of a charged particle in an arbitrary static magnetic field which possesses midplane symmetry. 2. A Taylors series solution about the central trajectory is then assumed; this is substituted into the general differential equation and terms to second-order~n the initial conditions are retained. 3. The first-order coefficients of the Taylors expansion (for monoenergetic rays) satisfy homogeneous second-order differential equations characteristic of simple harmonic oscillator theory: and the first-order
SYSTEM
AND
SPECTROMETER Y
DESIGN
79
Y
B x ~, (; ~ t Central trajectory, lies In magnetic midplane
/~
7
I/h = po
o
FIG. 4.
Curvilinear
coordinate
of equations
of motion.
dispersion and the second-order coefficients of the Taylors series satisfy second-order differential equations having driving terms. 4. The first-order dispersion term and the second-order coefficients are then evaluated via a Greens function integral containing the driving function of the particular coefficient being evaluated and the characteristic solutions of the homogeneous equations. In other words, the basic mathematical solution for beam transport optics is simila-r to the theory of forced vibrations or to the theory of the classical harmonic oscillator with driving terms. It is useful to express the second-order results in terms of the firstorder coefficients of the Taylors expansion. These first-order coefficients have a one-to-one correspondence with the following five characteristic first-order trajectories (matrix elements) of the system (identified by their initial conditions at t = 0), where prime denotes the derivative with respect to t: 1. The unit sinelike function SX(t) in the plane of bend (the magnetic midplane) where SX(0) ~ O;-S~(0) = 1 (Fig. 5).
80
K.
L.
BROWN
S.r(t)
Object
P()
Image trajectory
FIG.
5. Sinelike function
s,(f) in magnetic
midplane.
2. The unit cosine-like function CX([) in the plane of bend where c.(O) = l;c~(0) = O(Fig. 6). 3. Thedispersion function dY(r)int heplaneo fbendwhered.~(O) = O;d~(0) =O(Fig. 7).
FIG. 6. Cosinelike
function
{.T(1)in magnetic
midplane.
p. +
AP
Pu
-.
Fl~, 7. Dispersion
funclion
fi..(f)
in magnelic
midplane.
SYSTEM
AND
SPECTROMETER
DESIGN
81
plane where
Diane
Sy (t)
t
Object m Image
FIG. 8.
Sinelike function
s,(r) in nonbend
(y) plane.
plane where
Writing
the first-order
Taylors
expansion
position
of
i i i
i
an arbitrary trajectory at position t in terms of its initial conditions, the above five quantities are just the coefficients appearing in the expansion for the transverse coordinates x and y as follows: x(t) = cX(t)xO + sX(t)x~ + dX(t)(Ap/pO) and
y(t) =
C.(t)yo
S.(t)y:
where X. and y. are the initial transverse coordinates and x: and ,v: are the initial angles (in the paraxial approximation) the arbitrary ray makes
u
FIG. 9.
82
K. L. BROWN
with respect to the central trajectory. Ap/pO is the fractional deviation of the ray from the central trajectory.
1. TIIe Vector D[~erential Equalion of Motion
momentum
We begin with the usual vector relativistic equation of motion for a charged particle in a static magnetic field equating the time rate of change of the momentum to the Lorentz force:
P=e(Vx
B)
and immediately transform this equation to one in which time has been eliminated as a variable and we are left only with spatial coordinates. The curvilinear coordinate system used is shown in Figure 4. Note that the variable t is not time but is the arc distance measured along the central trajectory. With a little algebra, the equation of motion is readily transformed to the following vector forms shown below: Let e be the charge of the particle, V its speed, P its momentum magnitude, T its position vector, and T the distance traversed. The unit tangent vector of the trajectory is dT/dT. Thus, the velocity and momentum of the particle are, respectively, (dT/dT) V and (dT/dT)P. The vector equation of motion then becomes:
or -
$(:p)=ev($x B)
Pd2T m+#(#)=e(#xB)
where B is the magnetic induction. Then, since the derivative of a unit vector is perpendicular to the unit vector, d2T/dT2 is perpendicular to dT/dT. It follows that dP/dT = O; that is, P is a constant of the motion as expected from the fact that the magnetic force is always perpendicular to the velocity in a static magnetic field. The final result is: (1)
2. The Coordinate S~tem
The general right-handed curvilinear coordinate system (x, y, t) used is illustrated in Figure 4. A point 0 on the central trajectory is
SYSTEM
AND
SPECTROMETER
DESIGN
83
designated the origin. The direction of motion of particles on the central trajectory is designated the positive direction of the coordinate ~. A point A on the central trajectory is specified by the arc length t measured along that curve from the origin O to point A. The two sides of the magnetic symmetry-plane are designated the positive and negative sides by the sign of the coordinate ~. To specify an arbitrary point B which lies in the symmetry plane, we construct a line segment from that point to the central trajectory (which also lies in the symmetry plane) intersecting the latter perpendicularly at A : the point .4 provides one coordinate t: the second coordinate -r is the length of the line segment BA, combined with a sign (+) or () according as an observer, on the positive side of the symmetry-plane, facing in the positive direction of the central trajectory, finds the point on-the left or right side. In other \vords, .Y,~, and t form a right-handed curvilinear coordinate system. To specify a point C \\hich lies off the symmetry-plane, we construct a line segment from the point to the plane. intersecting the latter perpendicularly at B: then B provides the t~vo coordinates, t and Y: the third coordinate ~ is the length of the line segment CB. L;e now define three mutually perpendicular unit vectors (f, j, i). f is tangent to the central trajectory and directed in the positive tdirection at the point A corresponding to the coordinate t; .i is perpendicular to the principal trajectory at the same point, parallel to the symmetry plane, and directed in the positive .Ydirection. O is perpenciicular to the symmetry plane, arid directed away from that plane on its posi[ile side. -The unit vectors (i, j, f) constitute a right-handed system and satisfy the relations
The coordinate t is the primary independent variable, and we shall use the prime to indicate the operation (f/dt. The unit vectors depend only on the coordinate t, and from differential vector calculus, we may write .t = 171
j=o i =
11.c
(3)
84
K. L. BROWN
where /~([) = I/pO is the curvature of the central trajectory at point A defined appositive as shown in Figure4. The equation of motion may now be rewritten in terms of the curvilinear coordinates defined above. To facilitate this, it is convenient to express dT/dT and dcT/dTQ in the following forms:
The equation
In this coordinate
Differentiating
these equations
with respect to
t,it follows
that:
Tz = .YQ+ j2 + (1 +- i~x)~
1d - 2dt(
T 2 = xx + ~j + (i + /Ix)(~Ix + ~7X) )
hX)f
T = .fx + j] + (1 + and
+ (] + i?.v)t +
f(/?X
+ //x) for T
Using the differential vector relations of Eq. (3), the expression -. reduces to
T =
.f[X
i[2/~.Y+ /1.Y]
SYSTEM
AND
SPECTROMETER
DESIGN
85
The vector equation of motion may now be separated into its component parts with the result:
.f {
~),,,
(1 + I?x)(hx
/?x)] }
j, -
~;,),yx [-
.Vj) +
( I + Ax)(hx +
~x)]
[xx + JJj
B)
+ AX)B.Y .YB:] + f[xBv yBx]} (5)
Note that in this form, no approximations have been made; the equation of motion is still valid to all orders in the variables x and .Yand their derivatives. If now we retain only terms through second order in x and y and their derivatives and note that (T)z = 1+ 2hx + ~, the x and v components of the equation of motion become
x
(1 + hx)BV] (6)
j y(hx
The equation of motion of the central orbit is readily obtained setting x and v and their derivatives equal to zero. We thus obtain:
h = (e/PO) BY(O, O, t)
by
or
BPO = PO/e
(7)
This result will be useful for simplifying the final equations of motion. PO is the momentum of a particle on the central trajectory. Note that this equation establishes the sign convention between h, e, and Bv. 3. Expanded
Form of a Magnetic Field Having Median Plane Symmetr!
We now evolve the field components of a static magnetic field possessing median or midplane symmetry (Fig. 10). We define median
86
K. L. BROWN
Dipole
Quadruple
Sextupole
of magnetic rnidplane for dipole, quadruple, and sextupole The magnet polarities may. of course, be reversed.
plane symmetry as follows. Relative to the plane containing the central trajectory, the magnetic scalar potential q is an odd function in ~; i.e., ~(.Y,j, r) = V(.Y,~, t). Stated in terms of the magnetic field components B.,. Bu, and B!, this is equivalent tO saYing that:
B,(x, y, f) = B.(x,
and
j), f)
Bt(.Y, y, f ) = Bt(.Y, y, t )
It follows immed~ately that on the midplane B. = Bt = O and only Bu remains nonzero; in other words, on the midplane B is always normal to the plane. As such, any trajectory initially lying in the midplane will remain in the midplane throughout the system. The expanded form ofa magnetic field with median plane symmetry has been worked out by many people; however, a convenient and comprehensible reference is not always available. L. C. Teng(2) has provided us with such a reference. For the magnetic field in vacuum, the field may be expressed in terms of a scalar potential v by B = Vq. * The scalar potential will be expanded in the curvilinear coordinates about the central trajectory
* For convenience, wc omit the minus sign since we are restricting to static magnetic fields. the problem
SYSTEM
.4ND
SPECTROMETER
DESIGN
S7
lying in the median plane ~, = O. The curvilinear coordinates have been defined in Figure 1 where x is the outward normal distance in the median plane away from the central trajectory. y is the perpendicular distance from the median plane, f is the distance along the central trajectory, and h = h(t) is the curvature of the central trajectory. As stated previously, these coordinates (x, ~~,and t) form a right-handed orthogonal curvilinear coordinate system. As has been stated, the existence of the median plane requires that ~ be an odd function of ~, i.e., ~(x, ~, /) = p(x, y, t). The most general expanded form of ~ may, therefore, be expressed as follows:
(8) where the coefficients AQ~+~ . are functions of t. [n this coordinate system, the differential line element dTis given by
dT2 = dx2 + dy2 + (1 + hx)2(dt)2 (9)
1 22q +2+(l+hx)atla Zy
[
In [(l+hx)~ 1 o
flo)
Substitution of Eq. (8) into Eq. (10) gives the following recursion formula for the coefficients:
where prime means d/dt, and where it is understood that all coefficients A with one or more negative subscripts are zero. This recursion formula
88
K. L. BROWN
expresses all the coefficients in terms of the midplane field B,(x, O, t): where
A l.n 8BY
=
() dxn ~=o
functions oft
(12)
V=o
Since p is an odd function ofy, on the median plane we have B.. = Bt = O. The normal (in x direction) derivatives of BU on the reference curve defines BUover the entire median plane, hence the magnetic field B over the whole space. The components of the field are expressed in terms of p explicitly by B = Vp or
where Bt is not expressed in a pure power expansion form. This form can be obtained straightforwardly by expanding 1/(1 + hx) in a power series of Ax and multiplying out the two series; however, there does not seem to be any advantage gained over the form given in Eq. (13). The coefficients up to the sixth-degree terms in x and y are given explicitly below from Eq. (1 1).
hA~o Ala hA12 + h2A11 2hA; l 6h2A; o 6hhA;o A14 hA,3 + 2h2A12 2h3A,, 3hA;, 18h2A;1 18hhA; l + 36h2hA;o Als hA14 + 3h2A13 6h3A,2 + 6h4A11 A 50
=
(14)
A 51 = A~l 4hA~o 6hA% 4hA;o 11A;o+ 2A;3 6hA:2 2hAi2 + hA1z + 10h2A;l + 7hhA; l 411 zA11 J 3h2A11 16h3A~o 29h2hA~o + A1~ + 211A14 3h2A10 + 3h3A, z 3h4A11 (15)
SYSTEM
AND
SPECTROMETER
DESIGN
89
we can choose a circle with reference curve. The coefficients of the reference curve are then are greatly simplified by putting to zero.
4. Field Expansion to Second
radius A~m+ l,n in Eq. (8) and the curvature }1 all independent of t. Eqs. (14) and (15) all terms with primed quantities equal
Order Only
If the field expansion is terminated with the second-order terms, the results may be considerably simplified. For this case, the scalar potential q and the field B = VT become:
3nBU
ln=
ax ~=o
Y=o
and
A 30
=
[A;o
+ hA1l + Al,]
t. Then
where prime means the total derivative with respect to from which
B = V9
Bt(x, y>
t) =
1(1 + hx):
= (1 ;hx)[A~oy
+ ; xy l
(16J
By inspection it is evident that Bx, By, and Bt are all expressed in terms of Ale, All, and Alz and their derivatives with respect to t. Consider then BV on the midplane only BY(x, 0, t) = AIO + Allx + ~A12x2 + ..
dipole
- quadruple
sextupole
etc. (17)
90
K. L. BROWN
The successive derivatives identify the terms as being dipole, quadruple, sextupole, octupole, etc., in the expansion of the field. To eliminate the necessity of continually writing these derivatives, it is useful to express the midplane field in terms of dimensionless quantities n(t), ~(t), etc., or BY(X,o,
f) = BY(O, o, t)[l Mhx + ph2x2 + yh3x3 + ]
(18)
where as before h(t) = I/pO, and n, ~, and y are functions comparison of Eqs. (17) and (18) yields
of t. Direct
of motion
of the central
Combining become
+, A,2=
()
~]
~~] ]
(20)
To second order the expansions for the magnetic field components may now be expressed in the form: BX(X,y, t) = (Po/e)[ nh2y + 2~h3xy + B,,(x, y, t) = (Po/e)[h
-. 17/72x ~h3x2 + +(h
nh3 + 2~h3)y2 +
(21)
91
Fields
T = (BOr2/2a)sin 2a = BOx~/a
where BO is the field at the pole, a is the radius of the quadruple aperture and r and a are the cylindrical coordinates, such that x = r cos a and y = r sin a. From B = VP, it follows that
BX = BOy/a
and
By = BOx/a
(2~b)
Using the second of Eqs. (20) and Eqs. (22a) and (22 b).
?BV =BO = ]lJ12 P. ax ~=o () e a y=o
where B. is the field at the pole and a is the radius of the sextupole aperture. Using the third-pati of Eqs. (20) and Eqs. (24)
k: as follows:
(25)
These identities, Eqs. (23) and (25), are useful in the deri~ ation of the equations of motion and the matrix elements for pure quadruple and sextupole fields. 6. Tllc Equal io}ls of A!oti6n iti Tllcir Fi}lal Fornl to Secojld OrdiIr Having derived Eq. (21), we are no\v in a position t(>sljbstitllte int{} the general second-order equations of motion. Eq. (6). Combining
$
$ ~ ., ,: f
92
Eq. (6) (the equation of motion) with the expanded Eq. (21), we find for x x h(l + Ax) X(hx + hx) = (pO/p)T{(l and for y j y(hx + hx) +
hX)[-
/ : ? i
+ /rjj +
Note that we have eliminated the charge of the particle e in the equations of motion. This has resulted from the use of the equation of motion of the central trajectory. for T = (Y + 12+ Inserting a second-order expansion ( 1 + hx)2)12 and letting
(26)
+ +hx2
terms
(27)
/lh2}8
higher-order
terms
(~s)
for the
From Eqs. (27) and (2S) the familiar first-order terms may be extracted: X + (1 Il)h2.x =
116
equations
of motion
and
~ + )lhz) = O
(29)
Substituting k: = }lh2 from Eq. (23) into Eqs. (27) and (28). the second-order equations of motion for a pure quadruple field result b} taking the timit h -0, h ~ O and h ~ O. We find that .YJ+ k;x = k:xs l k:) = k:J8 where k: = (BO/a)(e/PO) = (Bo/a)( 1/Bpo) (30)
SYSTEM
AND
SPECTROMETER
DESIGN
93
Similarly, to find the second-order equations of motion for a pure sextupole field, we make use of Eq. (25) ~~13= k: and, again, take the limit /1-O, 11-0. and II ~ O. The results are: .Y + k:(x2 Jz) =
~?
2k:x) = o
where
(31)
of t}le Ta}lor s
The deviation of an arbitrary trajectory from the central trajectory is described by expressing x and y as functions of t.The expressions will also contain xO, YO,xL, jj and ~, where the SUbSCriPtO indicates that the quantity is evaluated at r = O; these five boundary values will have the value zero for the central trajectory itself. The procedure for expressing x and j as a fivefold Taylor expansion will be considered in a general way using these boundary values, and detailed formulas will be developed for the calculations of the coefficients through the quadratic terms. The expansions are written:
Here, the parentheses are symbols for the Taylor coefficients; the first part of the symbol identifies the coordinate represented by the expansion, and the second indicates the term in question. These coefficients are The ~ indicates summation over zero functions of t to be determined. K, ~, W, v, x: however, and all positive integer values of the exponents the detailed calculations will involve only the terms up to the second power. The constant term is zero, and the terms that would indicate a coupling between the coordinates x and y are also zero; this results from the midplane symmetry. Thus we have.
(Xll)=(yll)
=0
(33)
(xl
Yo)=(Yl~o)=o
(~1 Y6)=(Ylxb)=o
94
K. L.
Here, the first line is a consequence of choosing x~ = JO = O, while the second and third lines follow directly from considerations of symmetry. or, more formally, from the formulas at the end of this section. As mentioned in the introduction, it is con~enient to introduce the fottowing abbreviations for the first-order Ta}lor coefficients:
(x
I .Y~) = C.y(f)
(x
I -Y:) =
S.,,(?)
,: %
(J I yo) = cv(~)
(j Iy~)= s.(r)
(34)
Retaining terms to second order and using Eqs. (33) and (34), the Taylors expansions of Eq. (32) reduce to the following terms: c,.
.Y =
Sy
d,
+ (.Y\ X:)X8 + (.YI .Y~.Y:).Yo.Y: (.YI .Y08).Y03 + + (x I Joj:):o.l: + (.YI j;)2)j;)2 Su + (J I J:)J: (35)
and Cy .1 = (y [ Jo)jo
Substituting these expansions into Eqs. (27) and (28), we derive a di fferentiat equation for each of the first- and second-order coefficients contained in the Taylors expansions for .\- and .1. When this is done. a systematic pattern evolves, namely, c;. + k:.c., =
o
or
C;+
k:cY = O
S: + k~..$,. = O
s; + k;sv =
o
(36)
q:+ ~:q.,./:, =
q; + k;g!, = j;
where k; = (1 11)112nd k? = IIllz for the .Yand j motl~)ns, respeca tively. The first two of these equations represent the equations of motion for the first-order rnonoenergetic terms S.Y, .., s,,, and cu. That there arc c two solutions, one for c and one for s, is a manifestation of the fact that the differential equation is second order; hence, the tl~o solutions differ only bv the initial-conditions of the characteristics and (functions,
DESIGN
95
The third differential equation for q is a type form which represents the solution for the first-order dispersion dXand for any one of the coefficients of the second-order aberrations in the system where the driving term ~has a characteristic form for each of these coefficients. The driving function ~for each aberration is obtained from the substitution of the Taylors expansions of Eq. (35) into the general differential Eqs. (27) and (28). The coefficients satisfy the boundary conditions: c(o) = I
s(o) = o
d(0) = O g(o) = o
c(o) = o s(o) = 1
d(0) = O q(o) = o
(37)
The driving term f is a polynomial, peculiar to the particular 9, whose terms are the coefficients of order less than that of q, and their derivatives. The coefficients in these polynomials are themselves polynomials in h, h, . .. with coefficients that are linear functions of n, P,. For example, for g = (x I x:), we have
--
f=(2n-
1 #)h3c:
+ /lcXc; + +/?c:?
(38)
In Table 1 are listed the~functions for the remaining linear coefficient. the momentum dispersion d(r) and all of the nonzero quadratic coefficients, shown in Eq. (35), which represent the second-order aberrations of a system. The coefficients c and s (with identical subscripts) satisfy the same _differential equation which has the form of the homogeneous equation of a harmonic oscillator. Here, the stiffness k2 is a function of I and may be of either sign. In view of their boundary conditions, it is natural to consider c and s as the analogs of the two fundamental solutions of a simple harmonic oscillator, namely cos WI and (sin wf)/w. The function q is the response of the hypothetical oscillator when, starting at equilibrium and at rest, it is subjected to a driving force J The stiffness parameters k; and k: represent the converging powers of the field for the two respective coordinates. It is possible for either to be negative, in which case it actually represents a diverging effect. Addition of k; and k; yields
k:+k; =h2 (39)
-.
x) x) x) x) x) x) x) x)
DESIGN
97
For a specific magnitude ofh, A: and k; may be varied by adjusting n, but the total converging power is unchanged; any increase in one converging power is at the expense of the other, The total converging power is positive; this fact admits the possibility of double focusing. A special case of interest is provided by the uniform field; here h = const. and n = O; then k: = hz and k; = O. Thus, there is a converging effect for x resulting in the familiar semicircular focusing, which is accompanied by no convergence or divergence of y. Another important special case is given by n = +; here, k% = k; = h2/2. Thus, both coordinates experience an identical positive convergence, and Cx = CUand SX = SY;that is, in the linear approximation, the two coordinates behave identically, and if the trajectory continues through a sufficiently extended field, a double focus is produced. The method of solution of the equations for c and s will not be discussed here, since they are standard differential equations. The most suitable approach to the problem must be determined in each case. In many cases it will be a satisfactory approximation to consider h and n, and therefore k2 also, as uniform piecewise. Then, c ands are represented in each interval of uniformity by a sinusoidal function, a hyperbolic function, or a linear function off, or simply a constant. Using Eq, (36), it follows for either the x or y motions that: ; (es - cs) = o Upon integrating we find and using the initial conditions Cs cs = 1 on c and s in Eq. (37),
(40)
This expression is just the determinant of the first-order transport matrix representing either the x or y equations of motion. It can be demonstrated that the fact that the determinant is equal to one is equivalent to Liouvilles theorem, which states that phase areas are conserved throughout the system in either the x or y plane motions. The coefficients q are evaluated using a Greens function integral q = where G(f, T) = s(f)c(T) s(T)c(t) and q = ([) J t ~~(~)c(~) d7 C(t) dr (42) (43) J(~)G(t, Jo ~) d~ (41)
:/(T)S(T)
--
98
K, L. BROWN
To verify this result, it should be noted that this equation, in conjunction with Eq. (40), reduces the last of Eq. (36) to an identity, and that the last pair of Eq. (37) follows readily from this proposed solution. In particular, if~ = O, then q = O. Th~n it will be seen from Table I that several coefficients are absent, including the linear terms that would represent a coupling between x and y. Frequently, the absence of a particular coefficient is obvious from considerations of symmetry. Differentiation of Eq. (43) yields
t
,,
q = (f)
and
~f(7)~(~) ~T c(t)
t
f(T)~(T) / o
dT
(44)
q = ,f + $(t)
~ f(T)C(T) c(t) dT
f(T)~(T) dT
jo
The driving terms tabulated in Table 1, combined with Eqs. (43) and (44), complete the solution of the general second-order theory. It now remains to find explicit solutions for specific systems (>relements of systems. 8. Transformation
from Curvilinear Coordinates Coordinate System and TRANSPORT Notation to a Rectangular
All results so far have been expressed in terms of the general curvilinear coordinate system (x, y, r). It is useful to transform these results to the rectangular coordinate system (x, y, z), shown in Figure 4, to facilitate matching boundary conditions between the various components comprising a beam transport system. This is accomplished by introducing-the-angular coordinates O and p defined as follows (again, using the paraxial ray approximation tan 0 = 8 and tan ~ = q):
dy _ y
=~~=l+hx
(45)
where, as before, prime means the derivative with respect to /. Using these definitions and those of Eqs. (34) and (35), it is now possible to express the Taylors expansions for x, e, y, and ~ in terms of the rectangularcomdi nate system. For the sake of completeness and to clearly define the notation used, the complete Taylors expansions for
SYSTEM x, 0, Y, and
AND
SI)ECTR{)METER
DESIGN
99
x = (x iXO)XO + (x I do) 60
c;
s;
d:
e=(e + (6
x~)xo + (e I do)eo
x:)x:
+ (8
Y=
(Y IYO)YO
(y I qo)qo
Using the definitions of Eq (45), the coefficients appearing in Eq. (46) may be easily related to those appearing in Eq. (35). At the same time, we will introduce the abbreviated notation used in the Stanford TRANSPORT Program(3) where the subscript 1 means x; 2 means e, 3 means y; 4 means 0, and 6 means 8. The subscript 5 is the path length difference 1 between an arbitrary ray and the central trajectory. Rij will be used to signify a first-brder matrix element and Tijk will signify a
Im
K. L. BROWN
second-order matrix
form
.Y,=
$
)=1
R,jXj(0)
+~ ~
j=lk=j
Tij.Xj()~k(o)
(47)
where
.Y~ =
Using
x, X2 =
9, x3 =
-v, x4 = ~,x~=[,andx~=s
Eq. (45) defining d and p, the following identities among the \ arious matrix element definitions result: For the Taylors expansions for x we have:
Tlaz = (x T 134
=
y~j
(x
Y090) =
_T 14~ = (~
(48)
SYSTEM
AND
SPECTROMETER
DESIGN
101
(x I y:2)
(49)
expansion:
= (y I y,) = c,
I XOYO)
I xoqo) = (y I xojb) + /?(0)su
T 323 T 324
(y
(Y Y (Y
T 336 -( T 346
=
(50)
=( YIYO)=:(YIYO)=
c:
T 414 9 -(
q (P q (9
T 436 -( T 446
=
All of the above terms are understood to be evaluated at the terminal point of the system except for the quantity A(O)which is to be evaluated at the beginning of the system. In practice, /~(0) will usually be equal to h(~); but to retain the formalism, we show them as being different here. All nonlisted matrix elements are equal to zero.
102
K. L. BROWN
Matrix
Fornlalism
of Beanl Transport
Optics
The solution of first-order beam transport problems using matrix algebra has been extensively documented. (J-6) However, it does not seem to be generally known that matrix methods may be used to solve second- and higher-order beam transport problems. A general proof of the validity of extending matrix algebra to include second-order terms has been given by Brown, Belbeoch, and Bounin(7) the results of which are summarized below in the notation of this report and in TRANSPORT notation. Consider again Eq. (47). From ref. 3, the matrix formalism may be logically extended to include second-order terms by extending the definition of the column matrices xi and Xj in the first-order matrix algebra to include the second-order terms as shown in Tables IIV. In addition, it is necessary to calculate and include the coefficients shown in the lower right-hand portion of the square matrix such that the set of simultaneous equations represented by Tables IIV are valid. Note that the second-order equations, represented by the lower right-hand portion of the matrix, are derived in a straightforward manner from the firstorder equations, represented by the upper left-hand portion of the matrix. For example, consider the matrix in Table II; we see from row 1 that x = CXXO sXOO dX8 + second-order + + Hence, row 4 is derived directly by squaring follows: _ _ X2 =
=
(CXXO
terms as
C:X; +
S;8; 2SXdx908 +
+ d:82
The remaining rows are derived in a similar manner. If now xl = ~lxO represents the complete first- and second-order transformation from O to 1 in a beam transport system and X2 = ~zxl is the transformation from 1 to 2, then the first- and second-order transformation from O to 2 is simply X2 = ~2xl = MZMIXO; where Ml and ~2 are matrices fabricated as shown in Tables 11 and 111 in our notation or as shown in Tables IV and V in TRANSPORT notation.
SYSTEM
AND
SPECTROMETER
DESIGN
103
I
c
I f
{
TABLE 111
Formulation
of the Second-Order
Matrix
(y)
Plane
c,
Cy
Cxsy
SA CY
SXSY dxC, Sxc;
dxsy
Cxc; C*5;
etc.
00
:
h-
:
h c
00
00
92 Q
2
c .x .-
II
SYSTEM
AND
SPECTROMETER
DESIGN
107
III. Reduction of the General First- and Second-Order Theory to the Case of the Ideal Magnet
Section IIofthis reportwasdevoted to thederivationof thegeneral second-order differential equations of motion of charged particlesin a static magnetic field, In Section 11 no restrictions were placed on the variation of the field along the central orbit, i.e., h, n, and # were assumed to be functions of r. As such, the final results were left in either a differential equation form or expressed in terms of an integral containing the driving function~(t), and a Greens function G(?, T) derived from the first-order solutions of the homogeneous equations. We now limit the generality of the problem by assuming h, M, and ~ to be constants over the interval of integration. With this restriction, the solutions to the homogeneous differential equation [Eq. (36) ~f Sec. 11] are the following simple trigonometric functions:
SX(?) = (l/kX)
sin
kxf
(52)
where now
k: = (1 n)h2, k; = nhz,
and
h = I/Po
become constants of the motion. pO is the radius of curvature of the central trajectory. The solution of the inhomogeneous differential equations [the third of Eqs. (36)] for the remaining matrix clcmcnts is solved as indicated in Section 11, using the Greens functions integral Eq. (41) and the driving functions listed in Table 1. With the restrictions that kX and k, are constants, the Greens functions reduce to the following simple trigonometric forms: CX(t, ~) = (l/ky) and
G,(/UT) =
(l/k,,)
The resulting matrix elements are tabulated below in terms of the key integrals listed in Table VI, the five characteristic first-order matrix elements SX,CX,dX, c,, and .SY and the constants h, n, and P.
g
TABLE
VIa
Tabulation of the First- and Second-Order Matrix Elements for an Ideal Magnet in Terms of the Key Integrals
Listed in Table VIb
.r
R26
(~,1 2(2)1 1
(0
T,ll
= (0
.Y:)
~)h3]2,,
1 _
~c,,(t)c;(t)
T,,2
(0
.Yooo)
/lS:.(f)
p)h31212
+ +
T2,6
(8
T22,
(6
p)h31216
p)/131222 p)h31226 + +
T~,6
(0
0.8)
d..(f)]
SYSTEM
AND
SPECTROMETER
DESIGN
109
+1
111111
s +
m
I
e]
aaam
ww
--
11 II
II
II
11 II -
II
-11 II
II
II
II
II
II
II
II
II
II
II
II
Tabulation
The results are expressed in terms of the five characteristic first-order matrix elements sx(t), Cx(t), dx(t), c.(t), and 7 = O to Sy(t) and the quantities h and 11(assumed tp be constant for the ideal magnet over the interval of integration ~ = t). The path length of the central trajectory is r. From the solutions of the differential equations [Eq. (29) of Sec.
matrix elements
kx? ~lx(t) = (h/k;)[l
are:
Cx(l) =
Cos k.yf
cx(tj]
kut
110 =
~ [1
T) dT = ; l~x(t)
/ o
G.(f, T) dT =
t,
Ill =
Cx(T)Gx(f,
Jo
SYSTEM
AND
-..
----
. .
111
...
+
+
Icl
II
t
= ;2 -u [~,(t)
sy(T)Gv(t,T)
dT
tc,(t)l
4~; {~v(t)[l ~.v(t)l
.0
\
T) ~: _ (]T =
C.y(T)Cy(T)Gu(t,
1,
2k:s..(t).Y1,(t);
t J
2/t;s.,(t) Yu(t)
t
T) dT =
~: _ ~~;
rs~~=
{2~x(~)cY(~) ~,(t)[l
2 g
{[1 }
CX(T)~y(T)Gv(t,
1
+
$x(l)~y(t) Sv(t) + ~
.0
Cx(t)]}
.t
1
T)dT =
13Z3 =
k:
~o
~X(T)CY(T)Gy(f,
t 1
T)dT =
S,(f)[l
Cx(f)]
1~~4 =
k; 4k; {
h (133 ]313)
Jo
T)(jT =
~X(T)~Y(T)Gu(t,
13~~
Cy(T)f/X(T)Gu(t,
0 Jt
k;2
1314) = ~
kx
2k;sx(t)~y(f)} 1
ISAG =
T)dT =
Sy(T)dX(T)Gy(/,
Jo kx [ 2k;
k; 1 4k:
I
k; h (z34 k:
1
f2s.(f)cu(t)
Sy(r)[l + cx(~)l}
t
T)dT = s.(1)
140 T:.= @ = I
dt
Gy(l,
143
1~~= ~ ; ~Y(T)GY(t, ) (IT = ~ [sy(t) + try(r)] ~lf T I t 1 i~a = I;e = ~ sy(T)Gy(t, dT = j t~y(t)= ~33 T) dt ~o dt I 413 = 1:13 = Cx(T)sy(T)Gy(t, T) dT = k: ! 4k~ {(k: 2k!)~x(t)~v(t) k;s,(t)[l dt /o + c~(t)]}
Cx(t)]}
T)
I(lt 414
dT = k: ~2 2 $
cx(t)]
1314
dt T) dT = c,,(t)[ 1 +
x x Y 1
Jo t
_l 4k2 {()
CX(T)Cu(T)Gy(t,
1
_ 4k; {(k; 2k;)~x(~)Sy(t) cy(t)[l
1423 =
I;23
~ sX(T)Cy(T)GV(t, dt / o
T)
cx(t)cu(t)
k;sx(t)sy(t) }
+ ~
Cy(t)
x
14zh= I:Z4 = ~ dT =
k; 4h:
{
Sx(T)Sy(T)Gu(t,
dt /~ .
~+k;4k;{yy
1 ~/x( [ T) Gy(t, T)
dx(t)
1}
+ cx(t)] (k; 2k;)SX(f)cy(t)} 1
143~
Ig~~
dt
d.x(T)Gy(t, T) (]T =
o
; its,,(f) ~;~~
cy(T)
dT = $
~:
_
j c?,(t) + Sy(t)
~k~
dt
I 446
I;46
dt
~o
.$Y(T)
.
,
{(k:. Zk;).YA(f).$,(t) C,(t)[l -
Cx(t)]}
114
K. L. BRO\VN
The constants }Z~nd ~ are defined by the midplane field expansion [Eq. (18) of Sec. 11]:
For a pure quadruple, the matrix elements are derived from those of the general case by letting:~ ~ = O. k: = k; and k; = k:, where
k: = ]1112 (BO/a)(l/Bp) =
T 336 T 346
.? 43
T13G_= ~(kq/2)[kqt cosh k(,t+ sinh kqt] T 446 = ~kqt sinh kqt
(55)
SYSTEM
AN-D SI)EClROMETER
DESIGN
-115
2. Matri.~
Ele~?~e\~tsfor
For a pure sextupole, the matrix elements are derived from those of the general case by letting p)l = k; = (B,,a2)(l
/Bp)
k:t
~:[2 1~2 3 ~ ~t
= 1 R R;; -=t T 313 = k:t2 T 314 =Tst lh.2 3 T 323 =~~t lk2 3 T 324=6s ~k2t4 R 43
=0
=
R14 4
2k;t
k;t2
kft2 ~k?t3
(56)
116 3. First-
K. L. BRO\VN
a Cl{rre(i, Illclitled
, .,, -%, e
.,,, ~?! *%
Nlfitrix elements for the fringing iields of bending magnets have been derived using an impul~e ~ipproximati{~ll.y~) These computations, combined ~vith a correction term(g) to the R43 elements (to correct for the finite extent of actual fringing fields), have produced results ~vhich are in jubstarttial agreement with precise ray-tracing calculations and ~~i[h experimental rne:lsurements made on actual magnets. We introduce four ne~v \ariables (ii]ustrated in Fig. 11); the angle of inclination ~1 of the entrance Pace of a bending magnet, the radius-of curvature RI of the entrance fuce, the angle of inclination ~2 of the exit face, and the radius of cur~ature RJ of the exit face. The jign con~ention of~l and ,3: is considered positive for positi~e focusing in the transverse (j) direction. The sign con~ention for R, and R, is positive if the field boundary is convex out\\ard: (a positi~e R represents a negative sextupole component of strength. k~L = (// 2R) jecJ ~). The sign conventions adopted here are in agreement with Penner,(4) and Brown, Belbeoch, and Bounin.(7)
\ \> \ /, +
R2 /A
,//
Z1
,, y
,
\,
4 / po= I/h
Field boundaries for bending magnets. Definition of the quantities PI, ~~, Rl, and R, used in the matrix elements for field boundaries of bending magnets. The quantities have a positive sign convention as illustrated in the figure.
FIG. 11.
SYSTEM
AND
SPECTROMETER
DESIGN
117
The results of these calculations yield the following matrix elements for the fringing fields of the entrance face of a bending magnet:
R 33 1 RO = 34 T 313 = /7 tan2 PI
matrix elements are equal to zero. The quantity 41 is the correction to the transverse focal length when the finite extent of the fringing field is included.(g) ~~1= Kjlg sec ~1(1 + sin2 ~1) + higher order terms in (hg) where g = the distance between the poles of the magnet at the central orbit (i. e., the magnet gap) and
All nonlisted
BU(D)is the magnitude of the fringing field on the magnetic nlidplane at a position =. Gis the perpendicular (iistance measured from the entrance Pdce of the magnet to the point in question. B. is the asynlpt(>tic value of B,(z) well inside the magnet entrance. Tvpica] ~alues of h- ft>r actual magnets may range from 0.3 to 1.0 depending upon the detailed shape of the magnet profile and the location of the energizing coils.
118
K. L. BROWN
elements are:
R,l = 1 R 12 = 0 T 111 = (h/2) tan2 ~2 T 133 = (h/2) sec2 ~2 R21 = R 22 = T 211 = T212 = T 216 = T 233 = T 234 = R 33 R 34 T 313
= = =
l/fX 1
=htan
P2 h2(n
+~tan2~2)tan~2
h tan2 ~2
1 0
JI tanz ~z
R 43 = l/fv = h tan (F2 +2) R 44 = 1 T 413 = (J7/R2)sec3~2 +h2(2n+ sec2~2)tan 1414 = JI tan2 ~2 T 423 = h sec2 ~z T43, = Jz tan ~2 - hYzsec2(@z - @z)
P2
(58)
42 =
and K is evaluated
4. Matrix
Eiementsfor
a Dr\~t Distance
For a drift distance of length L, the matrix eleme~ts are simply as follows:
R,, = R,, = R~, = R44 = RJ5 = R,j = 1
SYSTEM
AND
SPECTROMETER Optical
119
IV.
Some
Useful
First-Order
II (1011)
may be reduced to a process of matrix multiplication. this is represented by the matrix equation
X1(l) =
$
J=l
R,,.Y,(o)
(j9)
where
that the determinant IR I = 1 results from the basic equation of motion and is a manifestation of Lioutilles theorem of conservation of phase space volume. The six simultaneous linear equations represented by Eq. (j9) may be expanded in matrix form as follows:
x(t) e(t? y(t) ~(t) l(t) 8(t)
where the transform ion is from an initial position 7 = L to a final pOsitiOn T = t. The zero elements R13 = RIA = Rz~ = RZ4 = R~l = RaQ = R.ll = R,lz = R3G = RqG = O in the R matrix are a direct consequence of midplane symmetry. If midplane symmetry is destroyed, these elements will in general become nonzero. The zero elements in column five occur because the variables x, e, ~. p, and 8 are independent of the path length difference 1. The zeros in row six result from the fact that we have restricted the problem to static magnetic fields, i.e., the scalar momentum is a constant of the motion. We have already attached a physical significance to the nonzero matrix elements in the first %ur~ows in terms of their identification with characteristic first-order trajectories. We now wish to relate the elements appearing in column six with those in row five and calculate both sets
1
R ~1
R,l
Rlz
O O R,, R,, O
0 0 R3, R,, 0
0 0 O O
x~
R,z
e.
Yo To
0 0
(60)
R ,1
R,,
/0
80
000001,
120
K. L. BROWN
in terms of simple integrals of the characteristic first-order elements cX(t) = Rll and s.(t) = Rlz. In order to do this, we make use of the Greens integral, Eq. (43) of Section II, and of the expression for the differential path length in curvilinear coordinates
dT = [(d.Y)2 + (dy)2 + (1 + hx)2(dt)2]2
(61)
of the equation
of motion.
Dispersion
The spatial dispersion dx(t) of a system at position t is derived using the Greens function integral, Eq. (43), and the driving term ~ = h(T) the dispersion (see Table 1). The result is for
dX(t) = Al, =
ST(f)
Jo
-t
Cx(T)h(T)
dr cX(t)
j0
S..(7)/?(7)
d7
(62)
where T is the variable of integration. Note that h(7) dT = da is the differential angle of bend of the central trajectory at any point in the system. Thus first-order dispersion is generated only in regions where the central trajectory is deflected (i.e., in dipole elements.) The angular dispersion is obtained by direct differentiation of dY(t) with respect to t;
t d:(t) = R2, = ~i(f) j o cx(T)h(T)
d7 cj(t)
f sX(~)/?(7) d~
j 0
(63)
where
c:(t) = R21
and
s~(r) = R22-
2. First-Order
Path Length
The first-order path length difference is obtained Eq. (61) and retaining only the first-order term, i.e.,
t
by expanding
/lo=(T
t)=
/ o
X( T)h(T)
from which t
/=x. t t
Jo
= RSIXO
cX(T)h(T) dr + 60
j0
+ R5260 + 10 + R56~
sX(7)h(7)
dT
/0
8 j 0
dy(T)/l(T)
dT
(64)
SYSTEM
AND
SPECTROMETER
DESIGN
121
Inspection
of Eqs. (62)(64)yie1ds
A. Achrornaticity
A system is defined as being achromatic if dy(t) = di(t) = O. Therefore it follows from Eqs. (62) and (63) that the necessary and sufficient conditions for achromaticity are that
t J
0
s-y(T)h(T)dT =
t J
0
cx(T)h(T)dT = o
(65)
Bycomparing Eq. (64) with Eq. (65), wenotethat ifasystemisachromatic, all particles of the same momentum will have equal (first-order) path lengths through the system. B. Isochronicity It is somewhat unfortunate that this word has been used in the literature: since it is applicable -only to highly relativistic particles. Nevertheless, from Eq. (64) the necessary and sufficient conditions that the first-order path length of all particles (independent of their initial momentum) will be the same through a system are that t jo
cX(7)h(7) d7 =
t t
j0
sy(r)h(7)
dT =
j0
dX(T)h(T) d~ = O
(66)
3. First-Order
Imaging
First-order point-to-point imaging in the x plane occurs when x(t) is independent of the initial angle 8.. This can only be so when
s.(t) = R12 = o
(67)
Su(t) = R~4 = o
(68)
First-order parallel-to-point imaging occurs in the x plane when x(t) is independent of the initial particle position Xo. This will occur only if
Cx(t) = Rll = o
122
K. L. BROWN
4. Magnl~cation
For point-to-point given by imaging in the x plane, the magnification
.Y(t )
is
M. =
and in the ] plane by
= Xo = 1~1,1 lc,x(~)l
(71)
M, = IR331 = Ic,(f)l
5. First-Order
lMotnentu~?z Resolution
For point-to-point imaging the first-order momentum resolving power RI (not to be confused with the matrix R) is the ratio of the momentum dispersion to the image size: Thus
R R, = R1~;o = dy(t ) Cx(t)xo
dx(t)
= c.%(t) SX(r)~(7)
o
d7
(72)
momentum
dY(t ) Cx(t )
=
dT
(73)
Jo
6. Zero Dispersion
For point-to-point imaging, using Eq. (72), the necessary sufficient condition for zero dispersion at an image is and
t J
0
o
Sx(T)h(T)
dT = O
cX(t) = O],
t !
Cx(T)h(T) d7 = O
(75)
AND
SPECTROMETER
DESIGN
123
It can be readily demonstrated from simple lens theory(4) that the physical interpretations of Rzl and RA3 are: c~(f) = RQ1 = I/fx and c~(r) = R+3 = l/~Y (76)
where ~X and fv are the system focal lengths in the .~ and y planes, respectively? between ~ = O and 7 = t.
From the results of Section III, we conclude that for an ideal magnet the matrix elements of R are simple trigonometric or hyperbolic functions. The general result for an element of length L is
cos k,L
1 sin~L
0 COSkxL]
kx
sin kxL
o cos kvL k,
sin
0 lsink ~y
COS
0 L 0 0 ;
x
R=
0 o
o 0 [kxL
sin
k,L
kgL
h sin k.xL E I
o-
- corkxL]
[
k.xL]
1
(77)
by letting
k: = k;
where
k: = /?lz2 = (BO/a)(l/Bp)
... +
COS k~L
k, sin k,L
R=
1
~
sin kqL
o o
cosh kQL
kq sinh kqL
0 0
# sinh k~L
0 0
O O
0 0 0 0 0 1
COS k~L
0 0 0 0
0 o 0
L
cosh kqL
0 0
and hyperbolic
0 0
functions
1 0
(7 8)
Note, that the trigonometric if the sign of BO is reversed. 9. The R Matri.~ Transfornled wil! interchange
to the Principal
Planes
The positions Z of the principal planes of a magnetic element (measured from its ends) may be derived from the following matrix equation:
100 Rlo
21
Oox
Oox 000 R,,
Xxxx
RPP =
001 0 0
100
xx 00
1 Z,x o
0100 0 0001 0000 0000
Solving this equation,
00
00
0 1 Zz, 00
00
10 01
tie have
1)/R21
0 (79)
21X = (Rzz
(80)
AND
125
magnet,
matrix
planes is
0 0 0
1
0 0 1 k, sin kyL
0 0 0 0
0 0
R PP
(82)
Correspondillgly, letting
R,,p is deri~ed by
and taking the limit 1?+ O for each of the matrix elenlents.
is: 1 I R k, sin kqL 0 0 0 0
The result
0
PP =
I I
@0
0
0
1
kq sinh k,L
0
1
0
0
0
0
o o
0
(83)
0 0
Z. = (l/k,) Z, = (l/k,)
0 0
tan (k,L/2)
0 0
0
1-
( 84)
-----
K. L. BROWN
A second-order aberration may therefore be determined as soon as a first-order solution for the system has been established, since the poly nomiai expressions for the driving termsj(~) ha~e ~ill been expressed tis functions of the characteristic first-order matrix elements (Table 1). Usually one is interested in knoyving the value of the aberration at an image point of which there are two cases of interest, point-to-point imaging s(l) = O and parallel-to-point imaging c(r) = O. Thus for point-to-point imaging,
..4 * ~ ~ j, .!
where r = { is the locution of an image and ~c~(f)l= ,M is the first-order spatial magnification at the image, and for parallel-to-point imaging,
where ~ = r is the position of the image and s(f) is the angular dispersion at the image. If a system possesses first-order optical symmetries, then it can be immediately determined if a given second-order aberration is identically zero as a consequence of the first-order symmetry. We observe that for point-to-point imaging a second-order aberration coefficient q will be identically zero if_the_product of the corresponding driving term J(7) and the first-order matrix element s(7) form an odd function about the
midpoint of the system. As an example of this, consider the transformation between principal planes for the two symmetric achromatic systems illustrated in Figures 12 and 13. We assume in both cases that the elements of the system have been chosen such as to transform an initial parallel beam of particles into a final parallel beam, i.e., Rzl = l/~X = Ofor midplane trajectories. We further assume parallel-to-point imaging at the midpoint of the system. With these assumptions, the first-order matrix transformation for midplane trajectories between principal planes is:
x(t) . x(t)
[
o 1
0
o
x(o) x(o)
8(0)
=
[
~(t) .
001
SYSTEM
AND
SPECTROMETER
DESIGN
127
B
FIG. 12. Three
bending
magnet
of
Thus c.y(~) = 1> ~.Y(~) ~ c~(f) = ~ ~(t) = 1 and f course = d.(r) = d~.(r) = O. About the midpoint of the system, the followlng symmetries exist for the characteristic first-order matrix elements and for the curvature ~~(~)= l/pO of the central trajectory; we classify them as being either odd or even functions about the midpoint of the system. The results are: /2(r) = even dx(T) = even s,,(T) = even cY(7) = odd j?(~) = odd d~(7) = odd s~(7) = odd c;.(7) = even
c~
/I
trajectory
PO
-.
Cx
imaging.
The principal
system with quadruple at center to achieve achromatic planes are located at centers of the bending magnets.
.,
128 As
K. L. BROWN
a consequence of these symmetries, the following second-order coefficients are uniquely zero for the transformation between principal planes.
This result is valid, independent of the details of the fringing fields of the magnets, provided symmetry exists about the midpoint.
I. Optical Symmetries in n = & Magnetic SJstems
In magnetic optical systems composed of n = j magnets having normal entry and exit of the central trajectory (i. e., nonrotated entrance and exit faces), several general mathematical relationships result from the n= ~ symmetry. Since k: = (1 n)hz and k: = nh, for n = 4 it follows that CX(7) = CU(T) and SX(T) = SU(T) at any position r along the system; thus as is well known, an n = ~ system possesses first-order double focusing properties. In addition to the above first-order results, at any point t in an second-order aberration n = ~ system, the sums of the following coefficients are constants independent of the distribution or magnitude (~h3) of the sextupole components throughout the system: (X I X:2) + (x I Yi2) = a constant
2(x I x;) + (Y I XOYO) =
independent
of ~h3
~h3
a constant independent of
(X I XOXi~ (Y I ~oYh) = a Gonstant independent of #h3 + (X I XO$)+ (Y I Y08) = a constant independent of ~h3
2(x I X:2) + (Y I x~y~) = a constant independent of ~h3
(X I x:8) + (Y I Y&8)= a constant independent of ~h3 (X I x3) + (X [y:) = a constant independent of (X\ XOX~) (XI YOYj)= a constant independent of +
Similarly,
~h3 ~h3
(85)
independent
of #h3
AND
SPECTROMETER
DESIGN independent independent independent independent independent of ~A3 of ~k3 of ~k3 of ~}t3 of ~A3
129
] X:2) + (~ I x~~~)
(86)
Of the above relations, the first is perhaps the most interesting in that it shows the impossibility of simultaneously eliminating both the (x I X:2)and (x I y~2)aberrations in an n = ~ system; i.e., either (x I X:2) or (x I yJ2) may be eliminated by the appropriate choice of sextupole elements, but not both.
VI. An Approximate
Aberrations
Quite often it is desirable to estimate the magnitude of various second-order aberrations in a proposed system to obtain insight into what constitutes an optimum solution to a given problem. A considerable simplification occurs in the formalism in the high-energy limit where POis very much greater than the transverse amplitudes of the first-order trajectories and where the dipole, quadruple, and sextupole functions are physically separated into individual elements. It is also assumed that fringing-field effects are small compared to the contributions of the various multipole elements. Under these circumstances, the second-order chromatic aberrations are generated predominately in the quadruple elements; the geometric aberrations are generated in the dipole elements (bending magnets); and, depending upon their location in the system, the sextupole elements couple with either the chromatic or geometric aberrations or both. We have tabulated in Tables VIIIX the approximate formulas for the high-energy limit for three cases of interest: point-to-point imaging in the x (bend) plane, Table VII; point-to-point imaging in the y (non bend) plane, Table VIII; and parallel-to-point imaging in the y plane, Table IX.
130
K. L. BROWN
TABLE VII
the Greens Function Solution, Eq. (22), in the High-Energy Limit as Defined Above for Point-to-Point Imaging in the x (bend) Plane, the Second-Order Matrix Elements Reduce to the Values Shown
Applying
TABLE
VIII
For Point-to-Point Imaging in they (nonbend) Plane, Eq. (23), the HQh-Energy Limit Yields the Values Given
DESIGN
131
For Parallel-( Line) -to-Point Imaging in the y (Nonbend) Plane, Eq. (24), the High-Energy Limit Yields the Values Shown
For the purpose of clearly illustrating the physical principles involved, we assume that the amplitudes of the characteristic first-order matrix elements CX,SX, dx, Cy, and Su are constant within any given quadruple or sextupole element, and we define the strengths of the quadruple and sextupole elements as follows:
L
/o
k: dr = k:L~
z ~
f,
,,
I
where Lq is the effective length of the quadruple, and where l/f~ = k, sin kQL is the reciprocal of the focal length of the gth quadruple; and for the jth sextupole of length L,, we define its strength as
L
;,
~,
I 1
k: dr = k~L, = S,
!,
The results are given in the tables in terms of integrals over the bending magnets and summations over the quadruple and sextupole elements. Note that under these circumstances the quadruple and sextupole contributions to the aberration coefficients are proportional to the amplitudes of the characteristic first-order trajectories within these elements, whereas the dipole contributions are proportional to the derivatives of the first-order trajectories within the dipole elements.
132
K. L. BROWN
As an example of the above concepts, we shall calculate the angle # between the momentum focal plane and the central trajectory for some representative cases.
For point-to-point imaging, it may be readily
tan+=
()
dx(i) 1 Cx(i) m)
= *)
f !
verified
that
Sx
da = (=
A,xo
(87)
where the subscript o refers to the object plane and the subscript i to the image plane. Let us now consider some representative quadruple configurations
and assume that the bending magnets are placed in a region having a large amplitude of the unit sinelike function SX(so as to optimize the first-order momentum resolving power R,).
Consider the simple quadruple configuration shown in Figure 14 with the bending magnets located in the region between the quadruples and s: ~ O in this region. For these conditions, ~1 = 11,SX= 11at the quadruples, and ~z = /3. From Table VII, we have:
where we make use of the fact that (/3/[1) = MX = cX(i). M. is the first-order magnification of the system. Hence, - _ t S. da tan~=.o / (88) (xl I Xj8) = a
fl
Sx
fz
Focal
-. P i I
FIGURE 14
SYSTEM
AND
SPECTROMETER
DESIGN
133
fl
/$
Focal
plane
FIGURE
15
2. Case II
For a single quadruple, Figure Ka/(1 15, the result + MX) is similar (89)
tan #.=
except for the factor K < 1 resulting from the fact that s, cannot have the same amplittie in the bending magnets as it does in the quadruple.
Therefore
i
SX da
K!la
3. Case III
Now let us consider a symmetric
In other words, the intermediate image has introduced a factor of t;vo in the denominator and has changed the sign of ~.
FIGURE 16
134
K. L. BROWN
References
for Magnetic Spcctrometcrs, HEPL Report NCJ. 104, High-Energy Physics Laboratory, Stanford University, Stanford, California ( [960). 2. L. C. Teng, Expanded Form of ,Magnctic Field with Median Plane, ANL
lntcrnal Memorandum 3.
K.L. Brom, D.C. Carey, Ch. Iselin, and F. Rothacker, TWSPORT A Computer Program for Designing Charged Particle Beam Transport Systems, SMC-91 (1973 Rev.),
NW-91, and CERN 80-04. S. Penner, . Calculations of Properties of lMagnetic Deflection Rell. Sci. /nsrr., 32, No. 2, 150-160 (1961).
4.
Systems, in
8.
9.
10.
and Texts in Physics and Astronomy, Vol. 17, R. E. Marshak, Ed.), lntcrsciencc, Ncw York, 1965. A. P. Ban ford, Tlle Trutlsport oj C/lurge~i Purtit /e Beat)l.s, Sp(~n. L[d.. London. 1966. K. L. Brown, R. Belbeoch. and P. Bounin, First-and Second-Order Magnetic Optics Matrix Equations for (hc hlidplane of Unif~)rm-l;ield \Vcdge ,Magncts. in Ret. Sci. /)ls[r.. 35, 481 ( 1964) (SLAC-PUB-19), Beam Optics of a Curvcci. Inclined R. H. HC]I1l, First- and Second-Order Magnetic Field Boundary in the lmpulsc Approxilll;]tl[)rl... SL,\C Rcp~>rt No. 24. Stanford Linear .Accclerator Center, S(anford, Ctilift)rnia ( 1963). the expression f[)r IJj and II,, w~s calcuA. C. Hindmarsh and K. L. Brown, lated in hnpublishcd work, at SLAC for inscr[ion into the TRANSPORT Program: K. L. Brown. Some Useful Thcorenls Based on a General First-Order Theory of Image Formation, SLAC lntcrnal Report, TN-64-18, Stanford Linear Accelerator Cen[er, Stanford, California.
--