The basic principles of the two-beam acceleration technique have been established in the first two CLIC test facilities, i.e. CTF [5.1], which is now out of operation, and CTF2 [5.2], which is still running. It is now proposed to demonstrate the overall feasibility of the many key issues which are specific to the CLIC scheme in two distinct successive stages.
The first stage, which would take five years, would be to build and exploit a new test facility (CTF3) which would demonstrate the feasibility, and test all the critical components of the RF power generation scheme albeit on a much smaller scale and with the drive linac at a different (higher) frequency. This facility would be housed in the present LPI (LIL+EPA) buildings.
The second stage, which would come immediately after CTF3 and which would take about five years, would be to build a limited, first-phase version (CLIC1) of the real CLIC power source to produce just one single-drive-beam unit rather than the multiple drive-beams it would ultimately be required to produce. This drive-beam would have the nominal CLIC energy and current, and would provide enough power in a ~ 624 m long section of the CLIC linac to accelerate a multibunch beam to 68 GeV. Since this is a final test of the CLIC scheme, all components will be definitive ones and, given a positive outcome of the test, would be used for the final construction.
Details of future CLIC studies and the new test facility (CTF3) are described in Ref. [5.3].
The technical feasibility of two-beam acceleration was first demonstrated in the CLIC Test Facility CTF1 [5.1] which was built to (i) study the production of short, high-charge electron bunches from laser-illuminated photocathodes in RF guns, (ii) generate high-power 30 GHz RF pulses by passing bunch trains through energy extraction cavities for testing CLIC prototype components, (iii) test beam-position monitors. A layout of CTF1 is shown in Fig. 5.1 . A 3 GHz 1.5-cell RF gun equipped with a laser-driven photocathode and operating at 100 MV/m produced a bunched beam with a momentum of 4.5 MeV/ c . A solenoid at the outlet of the gun provided some focusing of the beam before it was accelerated to 12 MeV/ c in a four-cell standing wave gun-booster cavity. Final acceleration to 65 MeV/ c was obtained using a 1 m long travelling-wave section -- provided by LAL. Energy was extracted from the beam by a 30 cm long travelling-wave section (CAS1) to provide short high-power 30 GHz RF pulses. This power was in turn fed to a second identical CLIC structure (CAS2) to produce high accelerating gradients. The decelerated beam then either went to a dump, or was turned through 180° by bending magnets at the end of the line and re-accelerated by the second high-gradient CLIC section. The facility was operated in either single-bunch or multi-bunch mode at a repetition rate of 10 Hz. Multiple bunches were made by splitting the laser pulse into a train of pulses each spaced by one 3 GHz wavelength. The synchronized laser system had been optimized at the fourth harmonic (262 nm) providing a maximum energy of 0.5 mJ per pulse (before splitting) and a pulse length of 8 ps FWHH. After an initial period of operation with CsI photocathodes, Cs2Te photocathodes were later used. These photocathodes were prepared in the laboratory and transported under vacuum and installed in the gun using a specially designed transfer system. The RF gun produced 35 nC in a single bunch and 450 nC in a train of 48 bunches. Only a small fraction of this charge, however, could be transported to the dump. The maximum 30 GHz RF power produced was 76 MW for 3 ns. The highest average accelerating field in the second CLIC accelerating structure was 94 MV/m.
Fig. 5.1 : Layout of the facility CTF1, which is now out of operation.
With the second CLIC Test Facility, CTF2 [5.2], a real two-beam accelerator, with separate drive and main beams, was built and successfully operated. A string of 30 GHz, low-impedance, power-extracting structures is used to decelerate the high-charge drive beam. The extracted 30 GHz power is transferred to a string of high-impedance structures accelerating the low-charge probe beam. A layout of CTF2 is shown in Fig. 5.2. The 30 GHz part of this facility is equipped with an active-alignment system with a few-microns precision. The 48-bunch 450 nC drive-beam train is generated by a laser-driven S-band RF gun with a Cs 2Te photocathode (PC). It is accelerated to 40 MeV average by two travelling-wave structures (TWS) operating at two slightly different frequencies to provide beam-loading compensation along the train. After bunch compression in a magnetic chicane, the bunch train passes through four Power Extraction and Transfer Structures (PETS) each of which powers one 30 GHz accelerating structure (CAS) (except the third which powers two) with 16 ns long pulses. The single-probe beam bunch is generated by an RF gun with a CsI+Ge PC. It is pre-accelerated to 45 MeV at S-band before being injected into the 30 GHz accelerating linac. The drive-beam RF gun produced a single-bunch charge of 112 nC and a maximum charge of 755 nC in 48 bunches. The maximum charge transmitted through the 30 GHz modules is 450 nC. A series of cross-checks between drive-beam charge, generated RF power, and main-beam energy gain have shown excellent agreement. A consistent set of values are given in Table 5.1 for two 30 GHz modules with one PETS feeding one CAS. The highest accelerating gradient obtained is 59 MV/m and the energy of an 0.7 nC probe beam has been increased by 55 MeV. Extremely high gradients [5.4] were obtained by powering a 30 GHz single-cell resonant cavity directly by the drive beam. The cavity operated without breakdown at a peak accelerating gradient of 290 MV/m. When pushed further, the cavity started to break down at surface-field levels around 0.5 GV/m. The breakdown manifested itself as a field extinction of the decaying pulse at different times in the pulse. At the end of the test, when the cavity was breaking down continuously, surface-field levels as high as 750 MV/m were obtained.
Fig. 5.2 : Layout of the present facility CTF2.
Since CLIC1 is a very large and expensive installation, a much smaller facility (CTF3) [5.3] is proposed as a first intermediate step to demonstrate the technical feasibility of the key concepts of this new RF power source, e.g., generation of interleaved bunch trains, operation with a fully-loaded drive-beam accelerator, and generation of accelerating gradients of 150 MV/m. The new CLIC Test Facility (CTF3) is shown in Fig.5.3. To reduce costs, CTF3 differs from the RF power source proposed for CLIC in the following ways (Table 5.2 ).
The frequency of the drive-beam accelerator is chosen to be 3 GHz instead of 937 MHz. This enables the 3 GHz klystrons, modulators, RF power compression units, and waveguides from the LEP Injector Linac (LIL) Complex to be used for power production which is always very costly. With 10 of these modulator/klystron units the drive-beam energy for a current of 3.5 A (~ half the nominal CLIC current) is 184 MeV -- this is very low compared to the 1.18 GeV for CLIC and obviously makes operation more difficult, but simulations indicate that it works. CTF3 only has the first two stages of the beam combination scheme, namely the times-2 Delay Line Combiner and the first Combiner Ring. The second (x4) large circumference Combiner Ring is very expensive and since it has the same scheme of combination it is not considered to be essential for this first demonstration test facility. The compression factor for the first Combiner Ring has, however, been increased from 4 for CLIC, to 5 for CTF3, to obtain an overall compression of 10. This gives a final bunch spacing of 2 cm (the same as in CLIC) for production of power at 30 GHz. Because of space limitations, it is unlikely that the circumference of the Combiner Ring can be made smaller than 84 m. This results in a final pulse length of 140 ns rather than the nominal CLIC value of 130 ns. The modulators produce a maximum RF pulse of 6.7 µs which after power compression with LIPS (x2.3) becomes ~1.6 µs. This beam pulse is long enough after a (x10) frequency multiplication to produce the required final 140 ns pulse. The drive-beam decelerator is limited to a total length of about 10 m (four transfer structures) compared to 624 m for CLIC. To limit the radiation produced by CTF3 it is proposed to run at 5 Hz instead of 75 Hz.
The new facility will be housed in the existing LIL and EPA buildings and use will be made of many of the LIL and EPA components. As mentioned above, an 84 m circumference ring appears just to fit in the EPA.
Fig. 5.3 : Schematic layout of nominal phase of CTF3.
The injector consists of a pulsed thermionic gun followed by one or two subharmonic (1.5 GHz) pre-buncher cavities, two 3 GHz pre-buncher and buncher cavities, and two 3 GHz damped/detuned accelerating structures (see section on drive-beam accelerator below). All the injector components sit in a solenoidal-focusing field. The final beam energy is about 26 MeV. The pre-buncher creates bunches with a spacing of 20 cm. Depending on the phase of the subharmonic pre-buncher, these bunches fall into either even or odd RF buckets of the 3 GHz system where they are trapped and accelerated. With the phase set at 0 degree, most of the charge (the process is not 100%) goes into bunches occupying only even 3 GHz buckets (in fact every other even bucket). As the phase of the subharmonic cavities is varied from 0 degree to +180 degrees, the intensity of the charge of the bunches in the even buckets is reduced and that in the odd buckets increased until there is charge in the bunches in the odd buckets only. This produces the drive-beam structure shown in Fig. 5.4. It is hoped that the phase switch can be done within 4 ns. A very broad-band power supply and a low- Q (~10) subharmonic buncher cavity are required in order to be able to switch in such a short time. This may not be feasible but it will only be known when the hardware has been tested. A longer switching time makes the system less efficient. The normalized emittance of the bunched beam at the exit (~26 MeV) is required to be <100 mm mrad with an r.m.s. bunch length <1.5 mm.
Fig. 5.4 : Bunch train structure.
An interesting alternative is to use a photo-injector consisting of a laser-illuminated photocathode in a high-gradient RF gun. This has the advantage of producing very short, low-emittance bunches. The bunch train structure could be generated directly by a suitable phasing of the laser pulses. The present CTF2 gun design with some modifications would be suitable for the RF gun, but the laser required for such a system has a specification which is beyond anything that exists today and an intense R&D programme would be necessary to determine if such a laser is indeed feasible and at what cost.
The 3 GHz drive-beam accelerator increases the beam energy from
~26 MeV to 184 MeV using 16 out of 20 normal-conducting TW structures
operating at 7.0 MV/m (4 structures being included in the drive-beam
injector with solenoid focusing). This linac has a conventional quadrupole FODO
focusing. To maintain beam stability in both this linac and the 26 MeV injector
linac the transverse wakefield levels have to be damped to
Q -values of 
100. This will be done by building new structures with waveguide damping
and detuning [the accelerating structures of the LEP Injector Linac (LIL) cannot
be used because of beam instabilities above 50 mA, and excessive beam loading
effects]. The structure design will be based on one developed for the CLIC 30 GHz
main-linac accelerating structures. The RF power is supplied by 30 MW klystrons.
After compression by a factor 2.3 by the existing RF pulse compression system
(LIPS) and splitting of the power, ~34.5 MW are provided at the input to
each 1.3 m long structure. Operating this linac in the fully-loaded condition
results in an RF-to-beam efficiency of ª 96%. Since
the bunch train charge is essentially constant along the 1.4 µs pulse, the
beam-induced energy spread is very small. A correlated single-bunch energy spread
is introduced in the drive-beam accelerator by a combination of off-crest running
and beam loading for BNS stabilization, and so that the bunches can be compressed
at a later stage. At the end of the accelerator there are ~2100 bunches of
2.33 nC per bunch, this corresponds to a current of 3.5 A averaged over
the train. The total single-bunch energy spread is expected not to exceed 1% approximately,
and the bunch-to-bunch energy spread must remain an order of magnitude smaller.
The length of the accelerating structure (1.3 m) has been chosen to give
maximum RF-to-beam efficiency with a current of 3.5 A when powered with 30 MW
klystrons. Since this is also the optimum length of structure for 4 A when
powered with 40 MW klystrons, it is proposed to replace 30 MW klystrons
when they fail with 40 MW klystrons so that there will be a natural progression
towards higher currents (4 A) and slightly higher energies (200 MeV).
The continuous train of bunches is split by the combiner delay line into a series of 42 m long bunch trains with 42 m gaps. It also produces a frequency multiplication (x 2) by interleaving the bunches in the even buckets with the bunches in the odd buckets to produce a bunch spacing of 10 cm. The two RF transverse deflectors in this line are short, 1.5 GHz, travelling-wave, iris-loaded structures whose fundamental mode is a deflecting hybrid mode. To prevent bunch lengthening, the lattice has to be isochronous. At the exit of the delay line the pulse current is 7 A.
A further frequency multiplication (x 5) is obtained in the 84 m
circumference combiner ring corresponding to a final bunch spacing of 2 cm.
This ring is more like a transport line than a storage ring because there is no
RF acceleration and the bunch trains make at the most a few turns (the first train
makes 4
turns and the fourth train
makes only half a turn). Since the CTF3 and CLIC frequency multiplication factors
are different, the way in which the trains are combined is also slightly different
(see the caption of Fig. 5.5).
Injection into the ring is made using a septum and two transverse RF deflectors (Fig. 5.5). The existing injection septum of EPA can be used for this purpose. The injection and extraction systems are located at a distance of half the ring circumference from each other. Beam loading and transverse wakefields in the 3 GHz transverse deflecting structures are a concern because the decrease of the transverse kick along the train will produce variations in transverse position of the bunches.
To prevent bunch lengthening, the combiner ring lattice has to be isochronous. In fact, since the fifth bunch train makes only half a turn in the ring, each half-ring also has to be isochronous. The suitability of using existing bending and quadrupole magnets of the LPI Electron Positron Accumulator (EPA) in the ring is being studied. The bunch length is kept relatively long in the combiners (~1.5-2.5 mm r.m.s.) to prevent the emission of coherent synchrotron radiation which increases the single-bunch energy spread as well as decreasing the average energy. The 140 ns bunch train is extracted from this ring using a pulsed kicker and a septum magnet both of which can be taken from EPA. At the ring-exit the peak current is 35 A.
Fig. 5.5 : RF deflector injection insertion
layout and lattice. Circulating bunches travel on the full-line orbits while the
injected bunches are kicked by the second deflector onto the
equilibrium orbit of the ring. The maximum outer trajectory (shown dotted and
intercepting the septum) is never followed, since the whole bunch train is ejected
before any of the circulating bunches reach the corresponding phase in the first
deflector. The maximum inner trajectory (also shown dotted) is not used either.
The bunch length will be tunable in the transfer line between the drive-beam accelerator and injection into the combiner ring so that coherent synchrotron radiation effects in the ring produced by the high-charge short bunches can be both controlled and studied. A design has been found [5.5] which incorporates this bunch-length tuning which is able to compress or stretch the bunches, in the strong-bending modules of the transfer line.
Owing to the unusual bunch energy correlation, the final bunch compression after extraction from the ring cannot be done with a simple three bending-magnet chicane because its optics does not have the correct sign of the R 56 matrix element, so a more complicated design has to be worked out which will certainly take up more space. The final r.m.s. bunch length for efficient production of 30 GHz RF power is ~0.5 mm.
The final 35 A drive beam will have an energy of 184 MeV and will consist of bunches with a charge of 2.33 nC/bunch with a spacing of 2 cm. It is proposed to use this drive beam either to power the four 30 GHz modules that have been built for CTF2, or to drive a high-power RF test stand (see later) for the testing of prototype CLIC components. To be able to produce sufficient power to generate the nominal CLIC accelerating gradient of 150 MV/m with only 35 A requires a slightly modified power extraction structure which couples more strongly to the beam, so four new power extracting structures will have to be built. After deceleration from an initial energy of 184 MeV to a final energy of 125 MeV the drive beam is sent to a dump. There is space in the modules for eight 86-cell 30 GHz accelerating structures with the potential to accelerate a beam from 150 MeV to 510 MeV. The main beam to probe the accelerating fields generated will be produced by a photo-injector which will be able to run in either a single-bunch or a multi-bunch mode. In the multi-bunch mode there will be about 50 short bunches of about 0.64 nC/bunch (the exact number of bunches will be determined by beam loading considerations). One of the two CTF2 RF guns can be used for the photo-injector but a new laser and pulse train generator will be required. The beam will be accelerated to an energy of ~150 MeV by four 4.5 m long LIL accelerating sections operating at a gradient of about 9 MV/m. When operating with the 30 GHz modules, a single-bunch main beam will be used because the constant impedance sections are not designed for multi-bunch operation.
It is proposed to operate a high-power RF test stand driven by the CTF3 beam in series with the four 30 GHz modules (see Fig. 5.3 ). The test stand would be a highly flexible experimental facility with a 1-2 m long test bed to measure a wide range of CLIC prototype components quickly, easily and accurately. This is in contrast to the modules where integration of prototype components into the very compact layout would always be problematic. The test stand would require instrumentation to allow precise measurements of beam energy spectra, beam profiles before and after interaction with RF structures, transverse wakefields (if feasible), RF power, dark-current spectra and X-ray emission.
It is proposed to use the LIL/EPA complex to carry out preliminary beam combination tests at very low currents as a first-stage CTF3. The transformation of the LIL/EPA complex into CTF3 would be carried out progressively. Recent studies and experiments have shown [5.6] that the EPA lattice can be modified (without making any hardware changes) to make it isochronous. It is foreseen to use the complex with small modifications to try out the (x 5) beam combination scheme. The idea is to combine five short (6 ns) pulse trains spaced at a distance equal to the circumference of EPA (~125.6 m or 420 ns) into a single 6 ns pulse. The total pulse train length is 2.1 µs. The initial bunch spacing will be 10 cm (3 GHz) and after the combination will be 2 cm (15 GHz). The combination will only be possible with very low pulse currents (0.3 A), the limitation being beam loading in LIL. The beam energy (~350 MeV) will be higher than that for CTF3 (~184 MeV) and should therefore make operation somewhat easier. To produce five useable pulses at a spacing of 420 ns will, however, require a major modification of the present LIL gun pulse-forming network. The new gun foreseen will be able to deliver seven pulses in order to take into account the transient beam loading in the structures and get five pulses for recombination into EPA. In the next stage, longer pulses and higher currents will be possible when the new CTF3 linac is installed. This will again require either major modifications to the present injector or a new injector. Operation of this first-stage facility would enable the RF deflectors and RF diagnostic equipment to be tried out and debugged, and should result in an earlier first demonstration of the overall scheme, albeit at a low current. The fast switching of the subharmonic buncher between even and odd RF buckets is only required when the delay line combiner is installed. Until this moment CTF3 will operate with one bunch every 3 GHz RF bucket and in consequence one-half of the nominal charge (the nominal current is always constant at 3.5 A).
Although quite a number of issues concerning the technical feasibility of the CLIC scheme can be addressed with CTF3, a full-scale prototype (CLIC1) to test one complete CLIC drive train will almost certainly be required before the community is convinced that the overall scheme will work. A layout of CLIC1 is shown in Fig. 5.6.
Fig. 5.6 : Schematic layout of CLIC1.
The drive-beam generator for CLIC described above will produce 22 drive trains per linac for a 3 TeV centre-of-mass collider. All major problems associated with this scheme can, however, be studied by generating only one drive train. To obtain the nominal beam energy with only 46 klystrons installed requires building forty-six 937 MHz RF power compressors (these compressors are not required for the final CLIC scheme). Klystrons working with a much reduced RF pulse (~25 µs) could be used at this stage instead of the CLIC nominal value of 100 µs. This first-phase CLIC installation would produce beams with the nominal current and would be able to accelerate a multibunch beam to 68 GeV at the nominal accelerating gradient of 150 MV/m. CLIC1 and CLIC (3.0 TeV) drive-beam parameters are compared in Table 5.2.
