Model Enhancements and Optimization of the Large-Signal Code Tesla

Summary form only given. During the last few years the large-signal code TESLA has been developed as a tool for the simulation and design of high-power klystron amplifiers. The model used in TESLA includes a self-consistent, non-linear solution of the three-dimensional equations of motion of electrons and the solution of the time-dependent field equations. One of the most important features in the model that makes the code fast and useful for design is the casting of the equations of motion with axial distance, rather than time, as the independent coordinate. However, this approach is valid only in cases where each electron\´s axial location is a monotonically increasing function of time. Once the RF field amplitude becomes large, some particle can stop or reverse direction of their motion. In this case the model in TESLA fails. To overcome this restriction we developed a new and separate treatment for these "special" particles which allows the code to be run in parameter regimes that were previously inaccessible. As an important part of the code improvement and development process we have continued to optimize TESLA to make it faster and use less memory. Features available in the Fortran-90/95 language were successfully implemented in the code. As a result, the code became more structured and compact; the amount of memory used was reduced almost by ~1 GB for the standard user\´s input. After some additional algorithmic optimization the code was also speeded up and became about ~15% faster than the previous version.