Parallel modeling of multiple beam klystons with tesla

Summary form only given. We present simulation results for a multiple-beam klystron using a new parallel algorithm in the TESLA code, where individual beams are modeled separately but coupled self-consistently through the resonators. Multiple beam klystrons (MBKs) are high power and high efficiency amplifiers that use separate beamlets, each propagating in its own beam tunnel, but interacting with common fields in gaps near cavities. In this way the perveance of the individual beamlets can be kept low, reducing space-charge effects, while the total beam current can be high, facilitating high-power operation. Mainly the beamlets are isolated from each other, except for their interaction with the common cavity fields. In the lowest approximation, the beamlets can be considered identical and the response of a single beamlet can be multiplied by the number of beamlets in the numerical simulation of a device. However, beamlets are not identical, having different R/Q values for example. Nevertheless, the high degree of isolation of beamlets allows for efficient parallel simulation of multiple beamlets. We have modified the large signal code TESLA by assigning each beamlet to a separate process and introducing MPI-calls to handle the communication between them through cavity fields. As the amount of communication is small, the running time of the code increases only slightly with beamlet number. Results for the effects of R/Q variations for the NRL MBK will be presented