Efficient calculation of multi-dimensional derivatives of various performance metrics of RF sources with respect to different design parameters is a critical element of their optimization and sensitivity analysis. The direct approach is to change slightly the value of a design parameter of interest and compute the resulting change in the metric of interest; an example is a calculation of how a small change in klystron cavity spacing affect output power. The major problem with this approach is a number of required runs of a simulation code. For example, when there are many (N) design parameters of interest then (N+1) runs are required. N can be very large for detailed design of RF sources for accelerators [*, **]. By computing the solution of the adjoint of the perturbed equations for the beam-wave interaction, we have shown [***] that all N partial derivatives may be computed with only three runs of the simulation code, no matter how large (N) is. Once calculated, these partial derivatives may be used to specify manufacturing tolerances and/or used in a design optimization calculation. We will also present examples of applications of adjoint approach to klystron and TWT design.

Klystrons and Multiple-Beam Klystrons (MBKs) are widely used or proposed to be used in accelerators as high-power RF sources. Development and optimization of klystron and MBK’s designs is aided by the use of different simulation tools, including highly efficient large-signal codes. We present an overview of capabilities of the TESLA-family of 2.5D large-signal codes, which have been developed at the Naval Research Laboratory (NRL) and which are suitable for the accurate modeling of single-beam and multiple beam klystrons. TESLA algorithm does support proper treatment of ‘slow’ and ‘reflected’ particles, what enables accurate modeling of high-efficiency klystrons. Recently developed more general TESLA-Z algorithm is based on the impedance matrix approach and enabled accurate, geometry-driven large-signal modeling of devices with such challenging elements as multiple-gap cavities, filter-loading, couplers and windows. Finally, recent introduction of the reduced-order, 1.5D versions of the TESLA algorithms enabled much faster, but limited modeling options. Examples of applications of TESLA-family of codes to the modeling of advanced single-beam and MBKs will be presented.

We present new capabilities in the Neptune electromagnetic particle-in-cell (EM-PIC) simulation code and design environment created to support geometry-based design of high power RF sources. Neptune’s time-domain EM-PIC model to simulate high-voltage, high-current electron beam/RF interactions is a key component of the first-principles design codes created by NRL and Leidos, which provide a comprehensive, geometry-based approach to RF source design*. Neptune allows importing multi-part device geometry created by conventional CAD tools, which can simplify the design process for complex 3D devices. Imported CAD parts can be manipulated, modified and combined with other geometric elements as needed using a constructive solid geometry (CSG) model to create the device geometry to be used for simulation. New features of the EM-PIC model include an improved waveguide port model, with time-resolved waveguide mode diagnostics, and support for customized electron beam models. We will summarize the new capabilities and present examples of applications to high power RF sources.