Room Temperature Accelerator Structures for Linear Colliders*

Early tests of short low group velocity and standing wave structures indicated the viability of operating Xband linacs with accelerating gradients in excess of 100 MeV/m. Conventional scaling of traveling wave traveling wave linacs with frequency scales the cell dimensions with λ. Because Q scales as λ, the length of the structures scale not linearly but as λ in order to preserve the attenuation through each structure. For NLC we chose not to follow this scaling from the SLAC S-band linac to its fourth harmonic at X-band. We wanted to increase the length of the structures to reduce the number of couplers and waveguide drives which can be a significant part of the cost of a microwave linac. Furthermore, scaling the iris size of the disk-loaded structures gave unacceptably high short range dipole wakefields. Consequently, we chose to go up a factor of about 5 in average group velocity and length of the structures, which increases the power fed to each structure by the same factor and decreases the short range dipole wakes by a similar factor. Unfortunately, these longer (1.8 m) structures have not performed nearly as well in high gradient tests as the short structures. We believe we have at least a partial understanding of the reason and will discuss it below. We are now studying two types of short structures with large apertures with moderately good efficiency including: 1) traveling wave structures with the group velocity lowered by going to large phase advance per period with bulges on the iris, 2) π mode standing wave structures. 1 HIGH GRADIENT RF BREAKDOWN The high gradient RF breakdown testing is reported in detail by Adolphsen [1]. There are several interesting and some surprising results that affect the choice of structure design, which we will discuss here. The first is that as suggested by Adolphsen the viable operating gradient appears to vary almost linearly with the inverse of the of the group velocity. A related observation is that the structures process very rapidly with a relatively small number of arcs up to a gradient that also varies slightly less than linearly with the inverse of the group velocity. Above that gradient the arcing rate increases dramatically and progress to higher gradients is very, very slow. The second and perhaps the most surprising result occurred during the simultaneous testing of a 105cm long structure and a 20cm long structure driven by the same klystron through a 3 dB coupler so that the drive levels and history would be identical. Both structures were designed to be approximately constant gradient, but precisely constant peak surface field on all disks. Both had an initial group velocity of 5% of the velocity of light. The short structure was identical to the first 20 cm of the long structure. One might have expected the 105cm structure to have roughly 5 times as many arcs as the 20cm structure. Instead, the two structures had equal numbers of arcs at all power levels within the statistical variation, except during the very early processing. This is less surprising in view of the fact that the vast majority of the arcs in the long structure occur in the first 20cm. The third interesting fact emerging from the high gradient testing is observed when a structure has been processed up to some level with a short pulse and the pulse length is increased significantly. The rate of breakdowns increases dramatically with the arcs distributed uniformly in time within the pulse, not concentrated in the added portion of the pulse. The fourth interesting result occurred in an experiment studying high electric field gradients in rectangular waveguide, Dolgashev [2]. The large dimension of the waveguide had been reduced to lower the group velocity to about 0.18c in order to raise the field strengths that could be reached with available power. The striking observation was that when the pulse length was less than 400ns and the peak surface gradient was 80 MV/m the arcs never degraded the high gradient performance of the waveguide. When the pulse length was more than 500 ns the arcs frequently degraded the high gradient performance. The degradation observed was a higher rate of arcing, or the inability to reach 80MV/m and higher xray levels on pulses where no arcing occurs. These four observations suggest that it may be important to consider the energy deposited at the site of an arc when an arc occurs for two reasons. First, it may alter the microwave parameters of the structure by causing a tiny, deposited-energy-dependent change in the resonant frequency of the cell in which the arc occurs and thus change the phase advance and the match of the structure. Secondly, there probably is a deposited energy threshold above which the high gradient performance of the structure is degraded. Above this threshold an arc is likely to cause surface damage which causes successive arcs to occur at or in the vicinity of the original site. For many years people have observed that RF processing is not always monotonic; that sometimes an arc causes a setback, lowering the RF power which the device being processed will accept. It has also been realized that it was ____________________________________________ *Work supported by the U.S. DOE, Contract DE-AC03-76SF00515. 0-7803-7191-7/01/$10.00 ©2001 IEEE. 3819 Proceedings of the 2001 Particle Accelerator Conference, Chicago