A new pulse compression system for intense relativistic electron beams

Nuclear Instruments and Methods in Physics Research 228 (1985) 217-227 North-Holland, Amsterdam

217

A NEW PULSE COMPRESSION SYSTEM FOR INTENSE RELATIVISTIC ELECTRON BEAMS D . TRONC CGR-MeV, Buc, France

J.M . SALOMÉ and K.H . BÖCKHOFF CEC, JRC CBNM Geel, Belgium Received 18 April 1984

A special 360° deflection magnet has been installed at the exit of the 150 MeV Geel Electron Linear Accelerator (GELINA) . It consists of five sectors with zero gradient fields and is designed to accept a 50% electron energy spread in the beam of the accelerator. This beam enters the magnet with a pulse width (fwhm) of typically 10 ns and leaves it with a pulse width (fwhm) of 0.6 ns . The peak current rises by this charge conserving compression from originally 10 A to about 100 A at the exit of the magnet . The compression is made possible by the time correlated electron energy degression caused by beam loading in the accelerator. The magnetic field

effectuates a phase space transformation, translating the energy dispersion into a time correlated spread of trajectory lengths. This results in a delayed arrival of the leading edge of the pulse at the exit of the magnet as compared with the corresponding arrival of the trailing edge, which means a pulse compression. The pulse shape transformation by the magnetic field and the conditions for ideal pulse compression are analysed . This is followed by a description of the magnet with its electron optics and by a communication about the results and the operational experience .

1 . Introduction

2. The pulse compression method

The pulsed beam of the 150 MeV Geel Electron Linear Accelerator (GELINA) [1] produces short bursts of neutrons in a mercury cooled uranium target [2], which are used in neutron time-of-flight experiments to study nuclear reactions as a function of neutron energy . Neutron energy resolution requirements in the keV and MeV range call for very short neutron and therefore also electron bursts . The shortest electron pulse widths obtained with S- or L-band linacs before the pulse compression system described below became operational were about 4 ns (fwhm) at maximum possible peak current, which in the case of GELINA amounted to 12 A. Shorter pulses were possible, however with reduction in peak intensity. Since the intrinsic potential of GELINA with respect to a further improvement of both parameters - intensity and pulse width - was fully exploited, possibilities of external pulse compression were studied. Pre-acceleration bunching with resonant systems turned out to be impracticable and costly . The use of a corresponding non-resonant system looked theoretically promising [3], the difficulties encountered however with its practical realization led to the study and finally the construction of the post-acceleration pulse compression system which is the subject of this contribution .

The electron pulses leaving the S-band accelerator consist each of a sequence of micropulses (in the following called bunches) which are assumed to have a width

0168-9002/85/$03 .00 © Elsevier Science Publishers B.V . (North-Holland Physics Publishing Division)

of the order of 10 ps and have, corresponding to the S-band frequency of 3 GHz, a time distance of 333 ps . For pulse widths considerably smaller than the filling time of the accelerator sections (1 .1 ps), the electromag-

netic energy stored in the cavities of the sections is the only source from which the electrons draw their energy . This is the case here since we deal with pulses of about 10 ns .

Each of the bunches of such a pulse (except the first one) finds on its way through the accelerator less stored electromagnetic energy available for acceleration than its forerunner because that one used already part of it . Consequently the total electron energy contained in a bunch is

stepwise decreasing with the sequence of bunches. For a rectangular pulse consisting of bunches of equal charge, the energy decrease should be monotonous with time and in the ideal case linear . It has to be noted here that each electron bunch has its own energy spectrum which may overlap with those of its neighbours . The electron energies in the spectrum of a bunch may also be time-correlated as in the classical textbook case shown in fig. 1 . There the first electrons

21 8

D Tronc et al. / A pulse compression system

E

Fig. 1. Electron bunches in phase stable position on the travelling wave E = Ec cos q) .

have the lower energies, the last ones the higher energies . In practice this should not be strictly true. The energy-phase relationship of relativistic electrons leaving the accelerator is generally more complex and can only be determined with larger computer programmes . An example of such calculations is given in sect. 2.3 . For the sake of transparency of the description of the pulse compression method we shall however assume here, and in sects. 2.1 . and 2.2 ., that the energy decrease from bunch to bunch as well as the energy-phase relationship inside a bunch are both linear . These assumptions will yield the conditions for ideal pulse compression . If we now deflect such an ideal beam in a magnetic field (in our case over 360°) so that the electrons are allowed to leave the magnet in the same direction as they entered it, then the time-correlated decrease of electron energies along the sequence of bunches in a pulse transforms into a time-correlated dispersion of trajectory lengths of the electrons at the exit of the magnet . The first bunch representing the leading edge of the pulse will have the longer trajectory and the last bunch the shorter one. Since all electrons have a speed very close to that of the speed of light, the arrival of the leading edge of the pulse at the exit of the magnet is delayed with respect to the arrival of the trailing edge : the pulse is compressed . With an appropriate design of the magnet ancf the vacuum chamber and with a sufficiently high vacuum in the chamber, the charge to the pulse will be conserved . The magnet effectuates a phase space transformation : the large intrinsic electron energy spread available is used to produce a small time spread . The inverse phase space transformation, translating a large available time spread into a small energy spread has been used at other accelerators [4] and has now become fashionable under the name particle beam cooling. The description of the method applied in our case is up to now very general. For a better understanding of the influence of the magnetic field on the pulses and bunches and of the intrinsic limitations of the pulse compression method more detailed considerations are necessary. These are given below.

2.1 . Transformation of the bunches by the magnetic field 2.1 .1 . Widths of the bunches Assuming that the beam is allowed to make a singleturn deflection in the magnetic field before it leaves this (fig . 2) and using the following denotations and dimensions: t, t' arrival times of a reference electron at [S] the entrance/exit, respectively, of the magnet, R (E) [m] radius of the trajectory, E, Eo [MeV] kinetic and rest energy, respectively, B [T] magnetic induction, c [m/s] speed of light, one may write for the time t' which an electron needs to reach the magnet exit : t'

= t + 277R/c = t + (27r/300 cB) E

since R=[E(E+2Eo)] i/z/(300B) .- E/(300 B)

for

E » E, .

From (1) we obtain for the finite width St, of a single bunch v at the exit of the magnet : Stv=St+KSE,=

1

[1+KI

dE)I di,

(2)

where St stands for the width of the bunches (which are assumed to be equal) before they enter the magnet . SE, is the electron energy spread to the bunch v and K = 2Tr/300 cB . SE and hence (dE/dt), will in general be negative as e.g . shown in the textbook case of fig. 1. By convention pulse and bunch widths are always counted positive. In principle the widths St .' of the transformed bunches

Fig. 2 Schematics of pulse compression. In the ideal case all bunches n,(t') are stacked one on each other at the same place, forming a compressed pulse with a width At' equal to St' and with an amplitude k times that of n'(t'). In this case St' = St' and ~', ~, = T' for all v.

21 9

D Tronc et al. / A pulse compression system

can be made almost equal to the widths St  by maximizing the induction B. Our goal is however to compress the pulses with a width d t which comprise a sequence of bunches, and not to compress the individual bunches themselves . In fact the width St' is very much larger than that of the original bunch St . The bunches become "debunched" by a magnetic induction which is not adapted to their compression but to the compression of the pulses . The leading edges of the bunches, representing the lowest energies, are largely overpassed by the trailing edges, representing the highest energies. 2.1 .2 . Shape of the transformed bunches

If n,(t')dt' are the number of electrons in the time interval dt' within the bunch v and n(t)dt the corresponding number in the time interval dt, then ny(t')dt=n  (t)dt,

provided no electrons are lost on their way through the magnet . From this and (1) and with E = E0 cos qp ; q) = wt, we obtain for the shape of a bunch transformed by the magnetic field n'.(t')=n  (t)/

1

+K

(dE)P]

(1)/(I - E,Kco sin wt).  =n

This relationship shows that the quantity 1 + K(dE/ d t ) , which stretches the width of the bunch v, reduces its intensity and transforms a rectangular pulse n  (t) into an asymmetric one. Maximum possible neutron output and at the same time smallest energy spread SE in a bunch of given charge are obtained, when the bunches are kept in the highest possible position on the wave, centered about phase 9' = 0. In this case half of the bunch has a negative phase. (This does not imply phase instability because the electrons are highly relativistic .) Expression (3) has a singularity at sin wt, = 1/EOKw . It means that at the time t 0 the leading edge of the bunch is dust being overpassed by the trailing edge (d t' = 0) and that then - theoretically - the intensity becomes infinite. For the parameters of our facility =120 MeV ;

E0

K

= 277/(300 cB) = 1 .75 x 10 -t0 [s MeV -1 ] ;

w = 1 .885 x 10 10 [s -i ] . This situation occurs at q?5 = 0.145° or i s = 0.134 ps . The shapes of the transformed bunches are shown in fig. 3 for two positions on the wave : (a)

(p0 =

-13.671 °

corresponding to (b)

BE

9p, = + 13 .671'

4), = + 13 .6711 = 2 x 3.4 MeV, T2 =

+ 41 .0150

~ ' n1

bt i =6t n =253ps 005 004 003

05-

002 001

~r

~6to =C500 ps-~i 6tjr 600ps

Fig. 3. Shape transformation of a rectangular bunch n P (l) into an asymmetric bunch n,(t') by the compression magnet for two positions of the bunch on the travelling wave . It is assumed that for both cases the charge density and the total charges are the same (St, = Si ll ) and that SE, = 3.4 MeV which corresponds to the measured value Sti = 600 ps . The figure illustrates the advantage of placing the bunches over the crest of the wave .

corresponding to BE = 26 MeV. The first of these cases is believed to correspond typically to the actual case. It reveals a sharp spike at cross-over condition. This could not be seen with the available measurement equipment. 2 .2 . Compression of the pulse

The pulses N(t) having the width d t consist of k bunches n P (t) with the widths St . Fig. 2 shows schematically how the pulses look like before and after the compression magnet. The.bunches n  (t) are well separated in time and have a period T = 1/f with f = 3 GHz. The bunches n,(t') are broadened, have a shape different from that of n (t) and must not a priori have a constant period r, . For any electron in the bunch n  (t) we may write according to (1) tl= t +

KE .

If we denote by t and t  +, corresponding instants of two subsequent bunches, e.g . those of the first electrons or those where the bunches reach half of the maximum amplitude, we get /P 

- 1,' =T,  =T

- K (E, -

E  ,) .

A necessary condition for a reasonable pulse compression effect is that

T.',+1 = ,r' =constant for all electrons in the subsequent bunches: the time-interval between the bunches should not depend on its position in the sequence. This means that E - E + , = AE, ,, = d E = constant . 

In other words : the energies of the bunches and of the

220

D. Tronc et al / A pulse compression system

electrons at corresponding places in the bunches must decrease linearly with time. In a specific experiment which was part of the feasibility study for the compression system an approximately linear relationship was demonstrated [5]. To achieve a constancy of T,+1 and with that the wanted linearity the following conditions have to be fulfilled: (a) the pulse width A t must be small against the fillingtime of the accelerator sections, so that only the electromagnetic energy Wem stored in the sections of the linac is used for the acceleration of the electrons. (b) all bunches must contain the same number of electrons, which means that the pulses N(t) must be rectangular in shape. Real pulses have a finite riseand fall-time. Linearity can therefore not be expected at the beginning and at the end of the pulses. (c) the functions (dE/dt)  in (2) which together with B determine the shape transformation of n (t) into n',(t') must be the same for all bunches in a pulse: the bunches must have the same position on the wave (constant phase) . This means that the widths St.' must be all equal. Otherwise no common period T' can be defined. If these conditions are fulfilled, one obtains ideal pulse compression by choosing the magnetic induction B such that T' = T -

297 0 or 300 cBa J E =

29r B° = .212JE[T ] . 300cTJE=0

(4)

In such an ideal case all the transformed bunches n .(t') are stacked one on each other on the same place, forming a compressed pulse N'(t') with a width At' equal to the width St' of the individual bunch and with an amplitude k times larger than that of the individual bunch n',(t'), where k is the number of bunches in the pulse. 2.2 .1 . Discussion Sensitivity of pulse compression. In practice the various intrinsic components contributing to deviations from linearity between the electron energy E and the time t will cause deviations from such ideal pulse widths . Independent of the effects of intrinsic imperfections it has to be especially noted that the pulse amplitude must be kept stable if one has fixed the value Bo for optimum compression, otherwise A E is changing and with that the slope of the function E(t) which means T' :* 0 and therefore pulse broadening . Multiple turn circular deflection [8]. A magnet which would allow such deflection would have the definite advantage that it can become smaller in two dimensions. The induction B can be increased proportionally to the number of possible turns. Correspondingly the width St . of the transformed bunches will be reduced

and with that in the ideal case also the width of the compressed pulses. For neutron time-of-flight experiments however pulses shorter than 150 ps are hardly useful since the neutron detectors have usually a larger time spread which dominates the time-of-flight resolution. In practice also the limitations mentioned above will in general not allow this compression potential to be fully exploited. 2.3. Model calculation of pulse characteristics In practice the bunch mean energies do not vary linearly with time and inside each bunch the electron energy versus time distribution does not follow any simple law. This is due to the bunching and beam loading processes . To consider these effects, a step by step simulation of the energy and phase evolution during the electron acceleration in the GELINA sections was carried out. This was done in the following way: a) The characteristics of a pulse leaving the first (standing wave bunching) section of GELINA, as they were calculated in the original beam dynamics study, were used to describe the pulse by 9 equidistant pseudo-bunches, consisting of 15 pseudo-electrons each . The energy E1 and phase (p, of each "electron" depend on the presence or absence of prebunching, of the accelerating fields and phases etc. [6]. b) Each of these pseudo-electrons crosses then the two following travelling wave sections. Denoting the three sections with 1, 2 and 3 the output energy is given by : E3 = E1 + d E2 cos(-P l +A012 ) + 4 E3 cos(¢l +A023) . AE, stands here for the maximum energy gain per travelling wave section and dp, i denotes the r.f. dephasing between sections i and j. We assume that d E, is proportional to the square root of the r.f. energy stored in that section, which depends itself on the beam loading by the successive bunches . An approximation of d E, is calculated for each bunch v by : JE, = E(W, -, - 4W,,j1/2 , AW, - ,áWR(W,, _ , )1/2 cos
221

D. Tronc et al. / A pulse compression system 403  = (297/X 0 )127rK( E3

-E)+ A tc v 9

5

1.

4 t denotes the pulse-width before compression . A characteristic result of the calculations is shown in fig. 4. It was obtained under the assumption of prebunching and with the input data :

A0,z = 0.3 rad, A023 = 0, A t = 10 ns, AW R = 0.127 . Although the shape details of the histogram are obscured by the poor statistics it reproduces surprisingly well the width of the measured pulse. Within the limits of the crude approximations the results indicate that the linearizations used in the simplified model were justified . 3. Description of the magnetic deflection system and its auxiliary equipment t Fig. 4. Histogram showing the shape of the compressed pulse following a computer simulation . The final phase with respect to a central "electron" which represents the fifth bunch out of the nine and which has the mean energy É is given by

Fig. 5. The magnetic deflection system .

The electromagnet * (fig . 5) has a total weight of 40 t and overall dimensions of 3.4 m X 3.1 m X 0.66 m. It is positioned horizontally with its vertical axis 8 m behind the exit of the accelerator and 4 m in front of the target . The horizontal dimensions are determined by the availa* Constructed by CGR-MeV, Buc, France .

222

D. Tronc et al. / A pulse compression system

INPUT PULSE

COMPRESSION MAGNET

Fig. 6. Schematic view of the compression magnet .

ble space in the target bunker . It was very fortunate that this bunker, when constructed in 1963, was made exceptionally large in order to have in the future also space for a magnetic deflection system to compress a 10 ns pulse of 120 MeV electrons down to 1 ns . The ideas how to realize the beam optics were at that time not available and no effort for the realization of such a system was made . The actual magnet allows a maximum equivalent

trajectory radius of 1.2 m (the curvature of a trajectory is not constant). Under the realistic assumption that the leading/trailing edges of a short pulse represent the respective electron energies 146 and 86 MeV, the time difference between leading and trailing edge of the pulse is shortened by the compression system by about 10 ns . Fig. 6 shows a schematics of the magnet with its sectors S, to SS [7]. The system has a symmetry axis perpendicular to the direction of the undeflected beam, so that S, is identical with S5 , S2 with S4 and S3 is split by this axis into two identical parts. In fig. 7 the beam-optical construction of the sector interfaces is illustrated for a given electron energy value: a) At A the electrons enter perpendicularly the sector face of S, The magnetic induction of S, is Bo/2 yielding a trajectory radius R . b) They continue their path with the same curvature through sector S, The exit face of S, which is identical to the entrance face of SZ is constructed in such a way that all electrons, regardless of their energy pass the interface SI/S Z perpendicularly. The radius r, the trajectory radius R and the angle a (see fig. 7) are linked by the relation tg(a/2) = r/R .

Fig. 7. The beam optical construction for a given electron energy .

c) The sector Sz , which together with the identical sector S4 constitutes the major part of the magnet volume, is designed for a magnetic induction Bo, thus twice that of S, and S5 . The centre of the trajectory for an electron with energy E is at CZ which is located half way between C, and B. d) The sector S3 produces again the induction Bo/2 . The trajectories for all energies have to pass the axis OY perpendicularly, which also means that the centres C3 of the curvatures are located on this axis . This condition

D. Tronc

et al. / A pulse compression system

determines the profile of the interface between SZ and S3 , which can be closely approximated by a circle of very large radius . The energy dependent point D of the profile lies then on C,C2 where C,C, = Cz D = R/2. The coordinates Xp and Yp of point D are then given by

223

a _ CROSS

Xo =2(C Z H-d)=Rsma-2d, Yo = AC, - C, H + C213 cos ß

SECTION

VIEW

3,43 m

= (R/2)(2-cosa+cos,8)

In these expression CZ H is the distance of Cz to AC,, d is the distance from A to OY and ß is half the angle of sector S3 (for the energy E) . e) The numerical values chosen for the realization of the pulse compression magnet are: d=0.426 m; r=0.586m and for the trajectory E = 146 MeV at Bo = 0.45 T: R XD

30OB-2 = 0.204 m;

.163m ;

a=30 .3° ; YD =

ß=6 .4°

2.304 m .

The above considerations are based on a geometrical construction of beam trajectories and magnet sector boundaries but not on an overall analytical description . In fact the relative simplicity of the calculation of this optical system is only apparent . A linear matrix analysis for a given beam .of large energy spread as it is usually applied for planar magnet boundaries is not possible here, since the sector boundaries S,/S, and S4/SS are circular and have relatively small curvature radii. They cannot be approximated by planar surfaces. Such a system belongs to a new family of circularly faced neighbouring magnet sectors. A further study of this family led us to the discovery that a magnet system with a two times 360° deflection can be realized but that an extension of the principle to a system with more than two turns for the electron beam is excluded [8]. Calculations were performed for six different energies from 86 to 146 MeV taking into account : a) The magnetic fringe fields measured for the pole distances e ( = 45 mm) (sectors with Bo = 0.45 T) and 2 e (sectors with Bo/2). The entrance and exit fringe fields depend on the free space left between the pole edges and the cylindrical shielding protecting against the side fringe field. The fringe field inside the electromagnet, near the interfaces of adjacent sectors depends on the 2 : 1 pole distance ratio, The shape of the free space between the pole edges and the cylindrical shielding ensures a variation of B from 0 to Bo/2, very similar to the variation of B from Bo /2 to B inside the gap of the electromagnet. The angle between the magnetic and the mechanical boundaries is negligible . The distance between the magnetic and the mechanical boundaries is equal to 0.29e . b) The slight radial increase of the magnetic field due

0

b_

MEDIAN

PLAN

VIEW

Fig. 8. The magnet design : (1) iron yoke ; (2) magnet coils; (3) vacuum chamber; (4) beam entrance pipe ; (5) beam exit pipe ; (6) pole gap 2e ; (7) pole gap e. to the decreasing losses in the iron yoke : (4B/B)/Ar - +1 .3% per m. c) The slight azimuthal decrease of the magnetic field in sectors SZ and S4 when moving from OY towards OX due to the absence of iron near the axis of propagation : (AB/B)/49- -0 .5% per radian . Azimuthal variations in sectors S,, S3 and SS are negligible. The computer simulations of trajectories make use of code TRAJ [6] rather than standard codes which need complete field mapping. TRAJ takes account of circular sector boundaries (planar faces can be approximated by choosing large radius values) and of fringe fields given by precise analytical approximations and their derivatives. Simulation of paraxial rays corresponding to a 16 mm diameter electron beam confirmed that the magnet acts in the trajectory plane as a free space and perpendicular to it in such a way that for a mean energy (which was 106 MeV in the GELINA case) : 1 .4 Yó )' (0 .6

2.0 Y, 1 .6)~ Y, )'

where Y and Y' denote the distances from the axis and slopes of the paraxial rays at the entrance i and the exit o of the magnet (in mm and mrad), respectively .

Fig. 9. Lay-out of the compression magnet and its location in the target room .

ACTIVATION ANALYSIS FACILITY

y

ó

á y5

5 n 0

ae

a

i 0 3 n

b

N N A

22 5

D Tronc et al. / A pulse compression system

Fig. 8 illustrates the magnet design with its coils, vacuum chamber, beam entrance and exit pipes, and pole gaps . The magnet and its position in the target bunker are shown in fig. 9. The vacuum chamber between the magnet poles is served by two 200 1 s - ' ionic pumps placed on the upper part of the magnet . Three other pumps of the same type are installed along the beam line between accelerator and magnet. Typical pressure obtained in the chamber is 4 X 10 -7 Torr without beam and 2 X 10 -6 Torr at full beam power. One of the vacuum gauges which is closest to the magnet is connected to the central pulse generator of the linac. In case the beam touches the vacuum chamber the pressure increases rapidly and a signal from that gauge reduces the pulse repetition rate of the beam injector by a factor of 16 . By this manipulation one avoids time consuming search for the beam after the vacuum has been reestablished. Two 20 mm diameter water cooled collimators, made of stainless steel are installed at 1 m and 3 m in front of the magnet, respectively, to ensure that only a well aligned beam of defined diameter can enter the chamber. 4. Results Measurements of the shape of the beam pulse are made with ferrite transformers placed in front and behind the magnet . Their rise time is about 1 ns, which

maan mouw

t

m

17H

Fig. 10. Typical results of beam pulse measurements with ferrite transformers before (a) and after (b) compression . Time scale: 5 ns/large division . Intensity scale: 12 A/large division.

11ns 9 A

Fig. 11 . Typical results obtained with a plastic scintillator, before (a) and after (b) compression . is not sufficient to obtain appropriate shape information on the compressed pulse. To get better information, in particular on the pulse width, a NE 110 plastic scintillator coupled to a RCA 8850 (2") photomultiplier has been installed at a backward flightpath of the linac. The scintillator sees that part of the bremsstrahlung-flash produced by the electrons in the uranium target, which succeeded to pass a 3.70 m thick water filled opening in the target bunker. The intensity reduction of the flash by this absorber is such that pile-up of detected y-quanta is avoided. On the average only about one out of hundred flashes yields a detector signal . The detector has a resolution of 300 ps . The shape of the compressed pulse is measured with a time-to-amplitude converter which has a resolution of 100 ps . Typical results of measurements with ferrite transformers are given in fig. 10, those obtained with the plastic scintillator in fig. 11 . The following observations can be made regarding these results: a) The pulse compression system transforms a pulse with a width of At = 10 ns (fwhm) into a pulse with an observed width of A10bs = 0.67 ns (fwhm), which, taken into account that the time resolution of the detector corresponds to a real pulse width of At = 0.6 ns (fwhm) . Since there are no observable beam losses, neither on the collimators nor within the vacuum chamber, the peak current rises from originally 10 A to a value of 100 A for the compressed pulse.

226

D. Tronc et al. / A pulse compression system

width BE of a bunch: SE

-200a 1 ns

a

r N 2

1

z

-1001

= at'/x = 3 .4 MeV.

The approximation is justified since St is almost two orders of magnitude smaller than JKSE.J. The value K = 1 .76 x 10 -10 [MeV - ' s] was calculated for an average induction of 0.4 T. From (4) we get for the energy-difference 4 E between two subsequent bunches, that means for the reduction of stored energy by one bunch, the value: 4 E = B o/0 .212 = 1 .9 MeV.

t Fig. 12 . The pulse shapes (a) and (b) (cf. figs. 3 and 4) obtained from computations are compared to an experimental result (c) (fig . 11). The three curves are normalized to the same charge. The value of the peak current was determined by fitting the measured shape to an analytical curve and normalizing the area under these curves to the measured charge in the pulse. A Gaussian shape fit did not represent the wings of the pulse in a satisfactory way. An excellent fit was obtained with the analytical form

N(t) =kt/11+(k2t)2 ] .

b) The compressed pulse shows a high degree of symmetry, which is very useful for the calculation of the neutron time-of-flight resolution function. c) Pulse-shapes following from two different approaches as discussed in chapters 2.1 ., 2.2 . and 2.3 . and shown in figs. 3 and 4 are compared in fig. 12 with the experimental pulse-shape of fig. 11 . Very good agreement is observed between the three curves after normalization to the same charge. The original, not compressed pulse, measured with the plastic scintillator, does not show a flat top as does the measurement with the ferrite core transformer . This is due to the changes in the absorption of the bremsstrahlung spectrum, resulting from the time correlated electron energy spectrum. The compressed pulse is not influenced by this effect since here the energy-time relationship is lost due to the superposition of the bunches. The compressed pulse measured with the ferrite shows some wiggles after the main pulse. These do not correspond directly to electrons in the pulse, as the comparison with the plastic scintillator measurement shows. They are probably due to waves excited in parts of the beam tube by the intense pulse. Relating the observations with the formulas derived in chapter 2.1, the following conclusion can be drawn: Assuming that the width At' of the compressed pulse equals the width St' of the individual bunch and neglecting in (2) 81 against IKSE 1, one obtains for the

The energy spectrum of an individual bunch therefore only overlaps with that of its next neighbours . 5. Operational experience To arrive at the best possible pulse compression, that means the shortest pulse width and highest peak current, various sets of linac parameters had to be tried. These investigations showed the importance of parameters such as the electrical field in the prebunching cavity of the accelerator and of the proper phasing of the different linac sections . They also showed the (expected) importance of the peak current stability for the pulse compression (sect. 2 .3). To improve this stability the quality of the power supplies for the focussing coils of the linac sections has been improved, likewise the temperature control for the cooling water of the sections . The profile of the compressed pulse was measured repeatedly over periods of 10 min with the plastic scintillator. The results always confirmed the observed pulse width of 0.67 ns . Extended measurements over periods of one day and more under normal running conditions - that means without specific care for peak intensity stability - yielded observed average pulse widths always shorter than 1 ns . It has to be remarked that due to the intersection of the magnet, dark current electrons cannot reach the target, which is important for some neutron physics experiments . The linac and the compression magnet have now worked together over several months without any problem, supplying to a rotary uranium target [2] electron peak currents of about 100 A with widths shorter than 1 ns. The average electron energy was 110 MeV under these conditions and the repetition rate 800 Hz. The authors appreciate the important contributions made by L. Rozenfeld during all phases of the realization of the magnet . They gratefully acknowledge the valuable assistance and cooperative efforts of staff members of CGR-MeV and of the GELINA team at all stages of the project. They express their gratitude to Dr . F. Poortmans for measuring the shape of the pulses with a plastic scintillator.

D. Tronc et al. / A pulse compression system

References [1] A . Bensussan and J.M . Salomé, Nucl . Instr . and Meth . 155

(1978) 11 . [2] J .M. Salomé and R . Cools, Nucl . Instr. and Meth . 179 (1981) 13 . [31 R .G . Alsmiller, Jr ., F.S . Alsmiller, J . Bansh and T .A. Lewis, Particle Accelerators, 9 (1979) 187 .

227

[41 H . Herminghaus and K .H . Kaiser, Nucl . Instr . and Meth . 113 (1973) 189. 15] J.M . Salomé and R . Forni, IEEE Trans. Nucl. Sci. NS-28 (1981) 2234. [61 D . Tronc, Internal Reports ST 7803 (1979), ST 8651 (1981) . [71 D . Tronc, French patent no . 7835 383 (1978) ; USA patent no. 4314 218 (1982) . [81 D . Tronc, French patent no. 8316 797 (1983).