NUCLEAR INSTRUMENTS AND METHODS 95 (t97I) I 3 I - I 3 5 ; © NORTH-HOLLAND PUBLISHING CO.

D I F F R A C T I O N R A D I A T I O N D E F O C U S I N G OF AN E L E C T R O N RING*

E. KEIL
CERN, Geneva, Switzerland

C. PELLEGRINI
Laboratori Nazionali di Frascati, Frascati (Roma), Italy
and
A. M. SESSLER
Lawrence Radiation Laboratory, University of California, Berkeley, California 94720, ~].S.A.

Received 23 February 1971

The influence upon axial stability in an electron ring of the criterion is obtained, and numerical examples show that the
diffraction radiation reaction force, generated by a ring moving criterion is not an important constraint upon the choice of
in an acceleration column, is calculated theoretically. A stability parameters or the operation of an electron ring accelerator.

1. Introduction We may obtain a rough estimate of the order-of-

It is well known that the diffraction radiation by an magnitude of the diffraction defocusing from a simple
electron ring in the acceleration column of an electron physical model. Consider a charge, Q, moving along
ring accelerator (ERA) is an important effect insofar as the axis of an acceleration column. The complete
it can cause significant loss of energy of the ring1'2). solution to Maxwell's equation is, in general, difficult
The effect of the diffraction radiation upon the internal to obtain, but roughly speaking there are image
dynamics of the ring has not so far been studied, charges moving in concert with the charge Q. These
although it is clear that the large energy radiation images are slightly displaced behind the charge,
could easily have a significant effect upon ring stability leading to an axial field, E,, at the charge and hence a
in the axial direction, where the f o c u s i n g - c o m i n g net retarding force. The magnitude of the displacement
only from ions, imagesa), and possibly from the is difficult to estimate. The gradient of this field,
accelerating field - is weak. which is what determines the focusing force, is,
In this note we study the contribution of diffraction however, not sensitive to the image charge displace-
radiation to the axial focusing forces of a ring, limiting ment. Thus, in a column of radius a, the field gradient
our analysis, for convenience, to the case of a ring d E J d z in the frame of the moving charge is approxi-
moving at relativistic speeds. We evaluate the defo- mately given by
cusing force for two different geometries: In section 2
we consider a charged rod and a current carrying rod (dEz)*
dz / a3 "
(1)
moving past an infinite array of semi-infinite perfectly
conducting plates, which geometry has the advantage Thus in the laboratory frame, dEz/dz is proportional
that the problem may be analyzed analytically. In to the relativistic v-factor of the charge.
section 3 we consider a charged ring in an accelerating The defocusing force of (1) will give a shift in the
column consisting of an infinitely long corrugated square of the axial oscillation frequency in the ring
cylindrical waveguide. The effect of the ring current frame, (o9~ y)2 of amount
is not included in this model. In section 4 we evaluate
the axial oscillation frequency resulting from defocusing
forces, and in section 5 we present some numerical a(°9*v)2 - move* k d z / (2)
examples.
where 09* is the revolution frequency in the ring frame
* Work supported in part by the U.S. Atomic Energy Com- and y* is the relativistic v-factor for the circulating
mission. electrons. This formula is derived in section 4, although
131
132 E. KEIL et al.

many readers may consider it obvious. Thus, from (1) where fl = v/c, and y = (1 _f12)-½, and P(2,y) is given
and (2): by eq. (34) of ref. 4.
The evaluation of (4), in the limit of y >> 1, follows
Av 2 ~ - U r e / ~ a, (3)
the procedure employed in section 3 of ref. 4. In
where N is the number of electrons in the ring, and r e particular, eq. (51) is modified to
is the classical electron radius. Taking N = 1013,
a = 10cm, and y ~ . = 4 0 - t y p i c a l parameters of an <E(~)> ~ ~
q (4f) x
ERA - we obtain Av2 = -- 7. × 10 -3, which is small in
comparison with the expected self-focusing.
We have, in this simple-minded discussion, ignored x Im
magnetic images which for a smooth accelerating
column would greatly reduce the Av 2. However, the
{( 1 - 2i d2exp(-B2)(l+z)
)fo ° } , (5)

with
structure of an accelerating column destroys the
nearly perfect electric and magnetic cancellation of a B= o -,1 (6)
smooth pipe and thus our result - obtained from con- Y
sidering only electric images - is a fair estimation of the
effect. i 2~
z = - + x/(2) (I +i) ((½) - - , (7)
2. Semi-infinite plates
and we have written tr = ao/~. The expression (5) is
In this section we consider as a model of an accele-
correct, in the limit of large Y, through the first two
ration column, an infinite set of semi-infinite conduc-
terms. Evaluation of the integral yields:
ting planes; i.e. a comb. The electron ring is replaced
by a charged rod and a current carrying rod moving
past the comb. The advantage of this model is that <E(~)> 2q
~ ? 2 ~ \ Xo / \ Xo/J
the defocusing f o r c e - j u s t like the radiation loss 4) -
can be calculated analytically. from which follows:
We employ exactly the notation of ref. 4, which
reference will have to be consulted to make the present q ~(½), 2nL'~,
<E(0)> = LY~ 2~7z \ Xo /
(9)
calculation understandable. The plates are taken in
the x - y plane and extend from - oo < y < oo, x > 0.
They are separated by the distance 2 nL, while the rod,
located at x = - x o, is parallel to the y-axis and moves d,~ /~=o 2xo~ ~\Xo/J"
in the z-direction with speed v.
The formula for <E(0)> shows that the average energy-
loss decreases as y - ¢ - which was the major result of
2.1. CHARGED ROD
ref. 4. On the other hand, the leading term in the
We first consider a rod having charge q per unit defocusing field varies linearly with y. It is easy to see
length. We want to compute the electric field in the that this leading term corresponds in magnitude 5)
z-direction due to the charges and currents on the to what one would expect from an image rod located
plates, but we only need this field Esz evaluated at at x = +Xo.
x = - Xo, z = vt+tr, and averaged over one period of We have numerically evaluated <E(tr)> from (4) for
the structure. F r o m eqs. (8) and (23) and the argument a number of values of p = 2 nL/xo and for y ranging
leading from eq. (23) to eq. (36) in ref. 4, it is easy to from 2 to 50. Taking p = 0.5, for y = 5 the asymptotic
see that formula (5) is only in error by 3%, while for y 1> 10
the error is less than 1%. For p = 3.5, the error is
<E(tr)> - < E , z ( - X 0 , z = v t + a , t ) > t =
about 7% at y = 5, but less than 1% for Y >/ 20.
{(?)( q )(1-i//~Y'~ x The numerical calculations are important for
evaluating how well <E(tr)> is approximated by its
= Im ~-L \1 +i/fly/
value and first derivative at ~ - - 0 . The calculations
showed that the diffraction fields (in contrast to the
x fo~ d;tP2(2, y) exp [ - 2 x° 2 + i ~ l ~ (4) self-field of a rod) were well-approximated by the first
Lr L _1)' two terms of a Taylor series over distances a < Xo/y;
 DIFFRACTION RADIATION DEFOCUSING OF A N E L E C T R O N RING 133

i.e. a o < Xo. In applications of this model to an ERA For the semi-infinite plate model, then, taking
we shall always satisfy this condition; i.e. the ring q = Ne/2nR, with R the ring radius
minor dimensions (in the ring frame) should be
smaller than the distance from the ring to the acceler- K - (g(t)), =
ating column wall. Thus, in a ring with non-zero minor
dimensions, which in the present model would be - 4NeZYrcRx~[1 + 3((½)~ \(2nL~½
l x o/ .3 [1 +(fl'Y)2] " (15)
approximated by a compact bundle of thin rods, the
field due to charges and currents on the plates is
adequately described by (9) and (10) with q corre- 3. Corrugated cylindrical waveguide
sponding to the total line charge of the ring. The self- In this section we represent the accelerating column
fields decrease as y-2 and we can safely neglect them by an infinitely long, periodically corrugated cylindrical
at large y. waveguide with geometrical parameters as shown in
fig. 1. We employ the notation of ref. 1 which is
2.2. C U R R E N T CARRYING ROD necessary for the understanding of what follows.
A rod having current in the y-direction of magnitude The complete vector potential A (r,t) is given as a
qfl'c is treated in appendix A of ref. 4. Employing sum over the eigenfunctions of the empty waveguide
Maxwell's equations to relate H~ to Ey one obtains A~(r):
( n ( a ) ) =-- ( H ~ ( - X o , z = vt+a, t)), = A (r, t) = ~ qz(t) Az(r), (16)
A

- 2qfl'Ylm{fo~d2p2(2,y)expV-2Xo2+i_~L2_l}" where the functions qz(t) obey the equation

L L L?
(11) #z+~o~qa = N~ 1 ~ j.A*dV=f~. (17)
dr"N
The leading terms in the focusing force are easily seen
N¢ is the number of cells and VN is their volume.
to be
If the azimuthal motion of the electrons is neglected,
an electron ring with charge Q and geometrical para-
\--~a/~,=o 2x 2 y 1+2---~y~ \ Xo / d " meters as shown in fig. 1, travelling with speed v, has
the current density
(12)
09
The energy loss, which is evaluated in ref. 4, varies as Jz ~ X
(fl'y)2y-L Since the transverse velocity of electrons, ~h(R~-R~)
in the ring frame, is approximately constant as the
x H ( ½ h - l z - v t [ ) H ( p - R O H(R2-p), (18)
ring is accelerated, the quantity fl'y is essentially
y-independent (and equal to unity, if the electrons have LI

!
relativistic transverse velocities before being accelerated
axially). Thus the energy loss of a charged rod and a
current carrying rod both vary as y -~ (at large y) and in
fact are equal in magnitude in this limit. In like manner,
the focusing force contributions [eqs. (10) and (12)]
become equal in the limit of large y. We believe this
equality to be a general (geometry-independent) result.

2.3. FOCUSING FORCE

The focusing force on an electron, in the axial []
direction, is given by
ea [OE~(a,t) fl' OHr(cr't)] (13)
F(t)= L ~-~ ~-~ .=o'
which we write in the form
Fig. 1. Geometry of a currugated cylindrical waveguide with a
F(t) = K (t) a. (14) charged ring.
134 E. K E I L et al.

where H(x) is the Heaviside step function. Performing Using (23), (24) and (25) we find the electric field
the integration in (17) with A~ from ref. 1 yields: gradient in the limit N~ ~ oo:

fz = _ Qv____~i~ AmS(½flmh) J(Zm) eXp(iflmVt). (19) lim t3Eza = Qdiogf2AtS(½ogah/v)j(o9a/yv) ×

Ne Ogz m Nc~oo ~Z z=vt

The factors S(x) = x - ~ sin x and J take into account × ~ AmflmS(½flmh) J(Zm)COSogztexp(-iflmVt ). (26)
m
the finite dimensions of the electron ring; J is given by
When this expression is averaged over the time
J(Zm) = 2 [ R 2 J I ( Z m R 2 ) - R ' J ' ( z ' R ' ) ] (20) necessary to traverse one period of the structure the
Zm(R22 " R~)Jo(Zm a) sum reduces to a single term and yields:

The propagation constants fl~ and Xm are defined in lim OEza '~ = Qdi [AtS(½ogah/v)j(og~/yv)]2"
ref. 1 by flo = oga/v-2 M/d with l chosen such that Nc-~o~ Oz z =vt/ 2 oga v
[flol <~ nld, fl,.= flo + 2 nm/d, ~m2 ---- O,)JC2 2 fl~.2 i
(27)
With the Arc cavities centred at z--O, and with j
from (18), f~(t) = 0 for It[ >1 T = ½ N¢ d/v, and hence Since At is not available in closed form, it is advan-
for t >~ T, q~(t) is given by tageous to compare (27) to the energy Ux radiated in
the 2th mode in one period of the structure, calculated

which becomes
qa(t ) = o9~-1
i7 fa(t')sinoga(t-t')dt', (21)
in ref. 1. We find

lim OEza ~ = ogzUa (28)

Qdi The total electric field gradient

qx(t) = - 2coZx ~ AmS(½flmh ) J(Z,.) ×

x {[S(49 +) + S(q~-)] sin o9~t + -~-~ Iz = o,/

+ [S(49 +) - S(q6-)] i cos o9~ t}, (22) is obtained by summing (28) over all modes. Because
where 49+ = ½ N~ d(oga/v+flm), and S(~b)= 49-' sin 49 of the factor oga it converges less rapidly as a function
as above. of ogz than the energy loss Ua.
For Arc~ ~ , the contributions to qa come from two Finally, we wish to remind the reader that in this
resonances with oga_+fltv = 0 in the notation of ref. 1. section we have neglected ring current effects.
In that limit we find:
4. Evaluation of the axial frequency
lim qa (t) = - Qdiog"~2 At S (½ ogz h/v) J (oga/yv) sin o9~ t. In order to evaluate internal ring dynamics it is
N c --*or)
(23) convenient to work in the frame of reference in which
the ring is at rest. In this frame, axial motion of
This result is multiplied by a factor of two because electrons is described by
two resonance conditions are fulfilled at the same
frequency by waves travelling in opposite directions d2z*/dt*2+og*2V2oZ* = F*(t*)/moY*, (29)
which are counted as one mode in ref. 1. where o9* is the revolution frequency, Vo describes the
The electric field gradient for the 2th mode is focusing due to ions, images, and the accelerating
OE~a c~Aza wave, 7" is the relativistic y-factor for the circulating
(24) electrons, and F* is the axial force on an electron due to
Oz z = , , = - ~a(t) ~---~1~=~,"
the diffraction radiation. The absence of a star on Vo
The z-derivative of the vector potential, averaged over follows from its invariance under Lorentz transfor-
the minor ring dimensions, follows from ref. 1: mation. We have neglected the change in energy of an
electron in the interval of an axial oscillation.
From (14), and the invariance of longitudinal force
OoAzZz= vt : - og-~l Z Am flm S (½ flm h ) J (zm)
under Lorentz transformation, we have

x exp (-iflm vt). (25) F* (t*) = K (t) a , (30)

 D I F F R A C T I O N R A D I A T I O N DEFOCUS1NG OF AN ELECTRON R I N G 135

which, since z* = aTN, becomes The corrugated cylindrical waveguide model for a
charged ring (ring currents ignored) yields
F * ( t * ) = K ( t ) z*/71l. (31)
B = 27rR~l/a 3 , (38)
In the case of relativistic axial ring velocity (711 >> 1)
and for closely spaced accelerating cavities, the time where the coefficient rl(b/a, d / a , g / a , 7) is a weak
variation of K ( t ) is rapid compared with an axial function of all of its arguments. Computations for a
oscillation period, and we may average K ( t ) over time. large number of cases indicate that r/ < 0.5. As
Thus letting remarked at the end of section 2.2, we expect that in the
relativistic limit ring current effects should introduce
K =- ( K ( t ) ) , (32)
an additional factor of 2 in ~/.
and combining (29), (31), and (32), we have Taking as typical values, N = 1013, R = 3 . 0 c m ,
7" = 40 one finds, from (36)
d 2 z */dt*2 + o9" 2 v 2 z* = O, (33)

where the total axial betatron oscillation frequency v v2 = v2 - 3 . 2 x 10 - 2 B . (39)

is given by
Thus (37) with x 0 ~ 1 0 . 0 c m and 711 >> I yields
~2 2 * *2 v2 = v 2 - 1 . 6 x 1 0 - 4 ; while (38) with R = 3 . 0 c m ,
= vo-K/(moT± ~o 711)" (34)
a = 10.0 cm and r/= 1.0 (to include magnetic effects)
We introduce the quantity B, by writing yields v 2 = vg - 6 . 0 x 10-4).
K = ( N e E 711~2fiR) B, (35) These defocusing effects are small, and presumably
can be easily overcome in practice by means of ion
where N is the number of electrons in the electron ring focusing or image focusing.
and R is the ring major radius. Clearly B has the
dimensions of inverse length squared. The factor 711
has been inserted merely for convenience. From (34) We are grateful to our colleagues in the Berkeley
and (35) Electron Ring Accelerator G r o u p for stimulating
v 2 = v 2 - N r e R B / ( 2 IrT*r) . (36) discussions and helpful comments. We wish to thank
Richard Hazeltine for having carefully checked the
In (36), r e is the classical electron radius and we have calculation of section 2 and, in particular, for cor-
employed ~o~ ~ c / R in deriving the equation. recting a previous error in section 2.2. One of us
The quantity B, upon which v2 depends, is a function (C. Pellegrini) is thankful for having been able to
of the geometry of the accelerating structure and of spend a year in Berkeley, during which term this
ring speed. For B > 0, the diffraction radiation present work was initiated.
reaction is a defocusing effect. Axial stability follows
if v2 is positive, and hence to obtain stability when
B > 0 requires that a non-zero amount of focusing
be supplied by ions, images, or the accelerating wave.
We have not concerned ourselves with radial motion
References
in this note as the focusing - from ions, images, and the
1) E. Keil, On the energy loss of a charged ring passing a
external field - i s strong in this direction, and there is
corrugated cylindrical waveguide, CERN Internal Report
no near danger of loss of radial stability. Crossing ISR/TH/69-49(1969); and in Proc. 7th Intern. Conf. Acceler-
of a resonance by a relativistic ring would - presum- ators (Yerevan, USSR, Sept. 1969) to be published.
ably - not be serious. 2) j. D. Lawson, Rapporteur's paper, ibid.
z) L.J. Laslett, On the focussing effects arising from the self-
fields of a toroidal beam, Lawrence Radiation Laboratory
5. Numerical examples Internal Report ERAN-30 (1969).
In this section we evaluate (36) for the structures 4) R . D . Hazeltine, M . N . Rosenbluth and A . M . Sessler,
discussed in sections 2 and 3. J. Math. Phys. 12 (1971) 502.
5) The sign is just wrong, however. The reason for this is that
The semi-infinite plate model, after replacing the
in the plate structure, since the plates are perpendicular to the
charge per unit length q by N e / 2 7rR and setting direction of motion of the rod, the boundary conditions are
fl'7 = 1, yields, from (15), satisfied, to fair approximation, by an image rod of the s a m e
sign as the rod (thus minimizing hrx along the plates); hence
B = (1/x 2) [ 1 - 0 . 7 7 8 (27rL/711Xo)~]. (37) the reversed sign in (dE/dtr).