CONCEPTION OF PHOTO-INJECTORS FOR THE CTF3
Laboratoire de l’Accélérateur Linéaire, IN2P3-CNRS, Université de Paris-Sud,
B.P 34, 91898 Orsay, France
In the framework of the CLIC Test Facility (CTF3) under development at CERN, LAL-Orsay is responsible for the construction of two photo-injectors for two different linacs. One, dedicated for the so called “drive beam linac” must fulfil very demanding specifications. The RF gun has to provide a high quality beam composed of more than 2,000 bunches containing 2.33 nC of charge each in one RF pulse. The model adopted is inspired from the CERN 3 GHz 2-1/2 cell type IV RF gun. We will summarize all the studies performed on the RF design and on the beam dynamics. The vacuum issue has been also carefully investigated. The constraints on the second photo-injector are less severe since it must be operated with one or 64 bunches of 0.5 nC each for the so called “probe beam linac”. It will also be a 2.5 cell gun at 3 GHz but its design will be substantially different with respect to the former. This last project has only recently begun and therefore we will show only the preliminary results of the RF and of the beam dynamics simulations. 1. Introduction
For 5 years, the LAL is involved in the CLIC-Test-Facility 3 (CTF3) under construction at CERN. The goal of this accelerator is to test and valid the concepts of the multi-TeV linear collider CLIC, proposed by the CERN . Schematically this accelerator is based on two linacs. One, the drive beam linac at 3 GHz, must provide a 150 MeV electron beam with high charge which passing through resonant decelerating structures should produce high RF power at 30 Ghz. Then, this power is sent to the 30 GHz travelling sections of the main linac where the electron beam is accelerated to the TeV energy for particle physics.
The drive beam linac of CTF3 was successfully operated with a thermionic gun built by the LAL [2,3] but it was decided to put a photo-injector in place of it because of the many advantages they have . This project is a part of a Joint Research Activity, called PHIN, itself part of the Coordinated Accelerator
*Research in Europe. Other work packages are defined in the context of an R&D program on photo-cathodes and the construction of a custom laser to drive the photo-injector. In addition, the LAL is in charge of the construction of a second photo-injector for the probe beam linac which is used as the main linac in CLIC. The rest of the linac should be built by a French institute, the CEA-Saclay in collaboration with the CERN. The first part of this paper is dedicated to the RF design and the beam dynamic simulation of the PHIN photo-injector. In a second part, the photo-injector of the probe beam will be described mostly on the aspects which differ to the previous.
2. The PHIN photo-injector
LAL owns a good knowledge in the fabrication of photo-injectors. The first one home made, CANDELA with 1,5 cell, was dedicated to accelerator physics . A second photo-injector for a small linac, ELYSE, according to a CERN type model, was built for radiolysis experiments in chemistry . Now, for the CTF3 experiment, we decided to benefit from the CERN experience with CTF2 . So, this gun is based on a previous proto-type CERN RF gun (2.5 cells, so called “type IV”) which was built and tested with beam at CERN . However, it needed further study in order to meet the specifications of the CTF3 drive beam linac (see table 1).
Table 1: Beam and RF gun parameters desired for CTF3.
RF frequency (GHZ) 2.99855
RF power (MW) 30
Beam energy (MeV) 5 – 6
Beam current (A) 3.51
Pulse train duration (µs) 1.548
Bunch spacing (ns) 0.6666
Charge/bunch (nC) 2.33
Repetition rate (Hz) 5
Bunch length, FWHM (ps) < 10
rms energy spread (%) < 2
< 25 Normalized emittance (：mmmrad) -10Vacuum pressure (mbar) < 2.10
In the following, we show the results of the RF simulations performed with a 2-D electromagnetic code, SUPERFISH, and a 3-D code, HFSS, which
* We acknowledge the support of the European Community-Research Infrastructure Activity under the FP6 “Structuring the European Research Area” programme (CARE, contract number RII3-CT-2003-506395)
allowed us to obtain the main characteristics such as the cavity dimensions and the axial field distribution. Beam loading is also evaluated. Then simulations of the beam dynamics in the gun with the PARMELA code are summarised. In addition the vacuum aspects have been thoroughly investigated, as in previous experiments with RF photo-injectors at CERN the pressure has been found to increase dramatically at high levels of extracted charge (?1 µC).
2.1 2D RF Simulations
The SUPERFISH simulations are used essentially to determine the physical dimensions of the photo-injector cavity. First, the main objectives are to check if the original design is compatible with the CTF3 parameters. We have studied, for example, the influence of the angle of the cathode wall as well as the shape of the iris. Of course, any change to the cavity must be compensated in order to keep the resonant frequency at the required value. The CERN design of the RF gun was optimised for higher charge by choosing the angle of the half-cell wall around the photo-cathode, to provide additional transverse focusing. According to electron beam dynamic simulations (see 1.3), this angle should be reduced to 3.4 ? rather than the previous value 8? in order to keep the energy spread within the desired limits.
Moreover, we have changed the shape of the iris from circular to elliptical to decrease the surface electric field and therefore it minimizes electrical breakdown and dark current levels . Furthermore, we introduced a slight asymmetry in the walls of the cavities for mechanical reasons and to try to reduce multipactor hazards. The new design is shown schematically in Fig. 1.
Figure 1: SUPERFISH design of the PHIN gun, lengths are in mm.
The evidence of the efficiency of the elliptical iris for the reduction of the surface electric field with respect to the circular one is illustrated in Fig. 2. At first sight, the gain is not obvious since the average field is roughly the same in the two cases. But the peak electric field, one the irises, in the elliptical design is 15 % lower with respect to the circular case. Now, the yield of the field emitted electrons by the cavity walls is strongly non linear with the electric field.
Therefore the use of the elliptical iris should be helpful for the reduction of discharge problems due to field emission.
02nd cavity1st cavity1/2 cavity
102030405060 segment Figure 2: Surface electric field as a function of the SUPERFISH segment on the circumference of the model in figure 1. The surface electric field is zero in the mid-cells and maximum on the iris. The circles show the elliptical iris case and the crosses correspond to the circular iris case. The axial electric field is 85 MV/m.
The parameters of this RF gun are shown in table 2. Finally, the main difficulty in designing the gun is to obtain good axial electric field “flatness” in every cell of the structure. This requirement is rather difficult to meet as the cells are strongly coupled and any change in the radius of one cell induces a variation of the electric field in every cell. The best adjustment we have obtained is shown in Fig. 3.
Table 2: RF photo-gun parameters Resonance frequency (GHz) 3.00305
6 Shunt impedance R (M，) s
Quality factor Q 14530
Figure 3: normalized axial electric field of the RF photo-gun.
2.2 Calculation of the Beam-Loading
It is well known that the electron beam interacting with the impedance of the RF structures induces a beam loading voltage which can be detrimental for gun operation . An important issue to consider is the HF matching of the photo-injector. The reflection factor depends on the coupling, (. In presence of
the electron beam, the coupling is given by:
(0(， (1) Pbeam1；PRF
where ( is the unloaded cavity coupling, P is the RF power dissipated 0RF
into the cavity walls and P = V*I is the power of the electron beam at the beam
output of the RF gun.
For the nominal current and electric field, the beam energy is 5.5 MeV hence P is 19.3 MW. The necessary RF power is 9.6 MW; therefore, to get ( beam
= 1 and no reflection in presence of the beam, the RF gun must be over-coupled with ( = 3. It also means we need roughly 30 MW of RF power to compensate 0
for the electron beam consumption. Besides the matching, the beam-loading induces a supplementary energy spread over the train of bunches. The beam loading voltage evolves as cavity voltage induced by the RF input power. Both grow exponentially with the time constant ~ = 2Q/? where Q and ? are LrLr
respectively, the cavity loaded quality factor and the resonant frequency. To avoid a decrease of the accelerating gradient over the train of bunches, one needs to inject the beam during the rise time of the electric field. In this way, the induced voltage is compensated by the increase of the RF input power resulting in a constant accelerating voltage. The time t of injection is obtained from the 0
evolution of the electric field due to the HF input power:
tE(t)，E(1？exp(？))max~ (2) E(t)0t，？~ln(1？)0Emax
with E(t), the accelerating peak electric field, e.g. 85 MV/m and E the 0max
electric field which would be established corresponding to 30 MW if no beam was injected. But without beam, the RF gun is not matched and the reflection factor is 25 %. So, only 22.5 MW goes into the gun which gives E = 130 max
MV/m. Using (2), one finds ~ = 0.385 µs and t = 0.408 µs. 0
2.3 3D RF Simulations
The RF power is transmitted to the photo-gun via wave-guides and enters the gun through coupling holes whose aperture must be adjusted to, in principle, minimise the reflected power. In our case, as explained in the previous paragraph, we must design the gun to get a coupling factor, (？ to be 2.9 which
will allow the gun to be at critical coupling in the presence of a beam at the nominal current of 3.51 A. Moreover, for beam emittance preservation, the electric field must be kept symmetric around the axis. Consequently we decided to connect two couplers symmetrically with respect to the horizontal plane. The 3-D design is shown in Fig. 4.
Figure 4: Half view of 3-D model of the RF photo-gun.
To leave enough space to install a solenoid around the gun it was decided to connect the couplers to the output cell and to use wave-guides whose inner height is 14 mm rather than the 3 GHz standard 34 mm. The shape of the coupling aperture is in the form of a racetrack and its dimensions are approximately 25mm x 10mm. This model was obtained after many simulations since three conditions need to be satisfied in simultaneously: the resonant frequency, the coupling and the field-flatness. For instance, according to the simulations it is necessary to have the radius of the last cell 1 mm smaller than in the second cell.
One last issue to check is the transverse symmetry of the electric field in the coupling cell. Thanks to the dual coupling, the field is symmetric vertically
with respect to the (x,z) plane. But, inevitably, there is a difference of the field amplitude between the x and y direction. It adds a quadrupolar component to the acceleration in the transverse plane which leads to a degradation of the emittance. One solution to counteract this problem is to have a completely symmetrical coupling, namely a coaxial coupling with a hollow cylindrical antenna going into the RF gun along the axis of the cut-off tube. This scheme is now becoming standard equipment on the photo-injectors dedicated for free electron laser experiments striving to reach ultra-low emittance of ~1 µmrad . But, in our case, we need a strong over-coupling which pushes the antenna, according to the simulations, deeply into the RF gun. The antenna would be practically at only 4 mm away from the iris between the middle cell and the coupling cell in an area of high electric field. In addition, the coaxial coupler has to support 30 MW of HF power. We considered all these features would enhance the breakdown hazards, hence we gave up this option. Nevertheless, it is possible to improve the electric field symmetry in the case of the dual waveguide coupling with a special design of the cavity. It was proposed by J. Haimson  to use a racetrack shape for the coupling cell in order to damp the asymmetry between both transverse directions (see drawing in figure 5a). Results of the HF simulations with this design are showed in figure 5b. First, it appeared that the field asymmetry is below 1 % in the case of the usual cylindrical shape. But, between 0 and 6 mm, where most of the particles are accelerated, the numerical noise, due to the finite number of tetrahedron, is dominating and makes difficult to analyse the results.
0.7Ez (arb. uni.)
051015202530x, y (mm) Figure 5: left, drawing of the section of the coupling cell; right, magnitude of the longitudinal electric field as a function of the distance in both transverse directions, red lines stand for the pure cylindrical cavity, green lines are for the racetrack cavity, plain lines represent the field along the y axis while the dash lines is along the x axis.
The goal is to reduce as much as possible the relative difference of the electric magnitude in the area of interest. With the racetrack shape of the cavity, if one looks far away from the axis at 25 mm, where the asymmetry is dominant with respect to the noise, the asymmetry is reduced by a factor 3. Then, it is very
likely that the field asymmetry is reduced in the same proportions in the space volume where the electron beam is present. The best reduction of the asymmetry, shown in figure 5, is obtained with a very small value of the straight part of the racetrack, only 0.6 mm is necessary.
2.4 Dynamics of the Electron Beam
*The simulations with PARMELA presented in this section are based on the
real longitudinal electric field calculated with SUPERFISH for the 2D model shown in figure 1. Unfortunately, it is not possible to extract the electric field of the HFSS model and to use it in PARMELA. Most of the simulations are done with 600 macro-particles for the study of the beam performances as a function of all parameters as the accelerating phase and gradient. Once the operation point is well defined, we will give beam performances with the maximum precision reachable with ten thousand particles. The results are often shown in the x direction because of the cylindrical symmetry and generally there are zero losses in the gun.
Several aspects of the beam dynamics have been investigated. First, calculations on the gun alone will be presented. Then, a study of the compensation of the beam emittance increase due to the space charge forces will be presented. The last simulations show the influence of the laser profile shape on the performances of the gun.
2.4.1 The natural behavior
The CERN design of the RF gun was optimised for higher charge, e.g. the choice of the angle of the wall around the photo-cathode. Therefore, we decided to study the influence of this parameter on the beam dynamics. The results are summarised in table 3.
Table 3: results of the simulations with PARMELA with 600 particles and the nominal current of CTF3 (3.51 A) as a function of the wall-angle around the photo-cathode. The widths are rms values and the rms emittance ； is normalised. T
Wall angle 0? 3.4? 8.3?
3.4 3.25 3.04 ； (mm) x
20.6 20.7 22.3 ； (：;mm-mrad) T
； (mm) 1.11 1.15 1.2 z
rms ！E/E (%) 0.6 0.75 2.6
* PARMELA from Los Alamos National Laboratory, modified by B. Mouton at LAL.
The case of an angle of 8.3? is incompatible with the gun specifications and the case of 3.4? seems less interesting with respect to a vertical wall. However, with a higher current, a small wall-angle is advantageous as it adds a focusing force which compensates the increase of the space charge forces.
So, as we want to maintain the possibility of operating at a higher current, we decided to adopt this small angle on the photo-cathode wall. Then we investigated the optimum operation point for the photo-injector as a function of the two HF parameters: the phase, ，, and the magnitude of the electric field, E.
As far as the phase is concerned, it was found that at 55? we have the maximum energy gain and the minimum emittance. At 35? the energy spread is minimized. We chose ， = 35? because it is more important to improve the energy spread than the little increase of energy gain at 55? and the slight degradation of the emittance can be largely compensated as we will see in next sections. As for the choice of the electric field magnitude, simulations showed that the emittance could be reduced by 10 % at 120 MV/m with respect to the nominal gradient, 85 MV/m. In addition the bunch length and the beam radius are smaller and the beam energy 40 % bigger. However, because of the beam-loading issue, 120 MV/m would require a power of 50 MW which is beyond the limit of the available HF generator and breakdowns hazards would be enhanced. So, the operating gradient will be 85 MV/m.
Results of the simulations for the gun model shown in Fig. 1 are illustrated in Fig. 6. The beam parameters at the output of the gun are summarised in table 4.
Table 4 : Beam parameters for a beam charge of 2.3 nC and a Gaussian laser spot ； = 1.4 mm (rms), r； = 4 ps and 1000 particles. t
E (MeV) 5.452
； (：mmmrad) 19.6 x
； (mm) 3.2 x
1.07, 3.56 ； (mm, ps) z
0.36 rms ！E/E (%)
0481216200z (cm) Figure 6: horizontal normalized emittance as a function of the longitudinal axis in the gun.
The comparison of Table 4 and Table 1 shows the performances of the photo-injector fulfil the specifications of the drive beam of CTF3. The electron beam dynamic is dominated by space charge effect. In Fig. 6, most of the emittance growth occurs in the first half-cell when the beam energy is below 1 MeV. Thanks to the magnetic focusing, it is possible somehow to reduce the emittance degradation.
2.4.2 Compensation of the space charge forces
Due to the space charge forces, the emittance grows linearly with the distance until the beam enters the accelerating section. At the output, the emittance is “frozen” because the space charge force is strongly damped at high energy. Therefore, to keep the emittance at the lowest possible value, it is necessary to use a transverse focusing between the output of the gun and the input of the section to compensate the transverse defocusing effect of the space charge. One possible technique, proposed by E. Carlsten , is the use of a magnetic lens. Simulations have been performed with a SLAC type section at the nominal current and 1.4 mm of laser spot size and with two coils for which the positions are variable (see figure 7). In all cases, thanks to a bucking coil, the magnetic field is less than one gauss on the photo-cathode. Besides, the fact that we used a SLAC section in the simulations has no influence on the conclusions which can be drawn from this study. For each couple of coil positions, we found the optimum value of the magnetic field for a good compensation of the emittance. Figure 7 shows that the biggest reduction of the emittance is obtained when the coils are placed near the photo-cathode. The only drawback is that it implies a stronger magnetic field, 0.25 T for the case where the coils are at z = 2 cm and 10 cm.
150200250300350400z (cm) Figure 7: Emittance as a function of distance for several kinds of magnetic field. The SLAC section is located at z = 107 cm and has a length of 63 cm and a gradient of 30 MV/m. The position of the coils is variable; square points, coil at z = 15 cm and B = 0.21 T; dashed line, 2 coils at z = 10 cm zand 20 cm and B = 0.23 T; plain line, 2 coils at z = 2 cm and 10 cm and B = 0.25 T. zz