A suite of 3D electromagnetic codes has been under development at SLAC for the past decade aiming to provide higher accuracy and model larger problems than existing codes through the combined use of unstructured grids and parallel processing. The codes are built on a unified data structure that allows them to share common pre- and post-processing capabilities. Wherever possible, they reuse existing parallel libraries such as ParMETIS [6] for partitioning and PETSc [7] for matrix-vector operations among others. Written in object-orientated C++, these tools are developed under industrial software engineering standards and use the Message Passing Interface (MPI) for communication on distributed systems. The code suite currently runs on NERSC's IBM/SP supercomputer as well as Linux clusters at SLAC and elsewhere.

• Omega3P (R. Lee, Z. Li, Y. Sun, R. Uplenchwar - SLAC) --
  The development of Omega3P began with the DOE Accelerator Grand Challenge and continued into SciDAC's AST project. It was motivated, nearly a decade ago, by the need for a high accuracy modeling capability to support the NLC accelerating structure R&D. Available eigensolvers were not able to predict the mode frequencies of 3D complex cavities to within 0.01% accuracy as dictated by design requirements [8]. Supported by both simulation and accelerator projects, ACD initiated a concerted effort involving the Scientific Computing and Computational Mathematics program (SCCM) at Stanford to develop a high performance, high accuracy eigensolver that can meet such a requirement.
• S3P (L. Ge, R. Lee, Z. Li - SLAC) --
  Now in its second year of development, S3P is a new finite element solver based on Omega3P to find the scattering matrix of open structures directly in the frequency domain. Before S3P, such calculations have to be done in the time domain using Tau3P which is only first order in spatial accuracy whereas S3P can easily be extended to higher-order bases. In addition, the computational kernel in S3P is solving a linear system so it also benefits from the ongoing research being done on linear solvers for Omega3P.
• Tau3P (C. Ng, A. Guetz, M. Wolf - SLAC) --
  While a large number of accelerator applications can be modeled in the frequency domain using Omega3P and S3P, important calculations such as pulse propagation and wakefields effects are best performed, or are only possible in the time domain. Tau3P is the 2nd in SLAC's series of parallel, unstructured grid codes which was initiated during the Accelerator Grand Challenge and reached full development under SciDAC. This time domain solver is based on the Discrete Surface Integral (DSI) method formulated on the generalized Yee grid to conform to curved surfaces for improved geometry modeling. The Tau3P challenges have been in suppressing late-time instabilities, mesh generation for complex structures, improving speedup, and the implementation of physics capabilities/boundary conditions that enable realistic simulations. Much progress has been made in all these areas so that a wide range of applications have been modeled successfully with great impact on accelerator projects such as PEP-II and the NLC.
• T3P (M. Kowalski, C. Ng - SLAC) --
  Development has also started in the past year on a new parallel time-domain code T3P that uses an implicit, unconditionally stable, Newmark-Beta time integration scheme to avoid the late time instabilities that had initially plagued Tau3P. This code is based on tetrahedral elements which allow for much easier meshing compared to Tau3P's hexahedral meshes, and open the possibility for prescribing beam transit along curved trajectories. Furthermore, the FEM formulation can be extended to higher-order bases in a straightforward way. Initial testing of basic functionalities in the code has been done and the focus has turned to the important development of wakefield calculation and S-parameter extraction capabilities. T3P can potentially replace Tau3P as the main time-domain solver upon completion of these implementations.
• Track3P (V. Ivanov, A. Guetz, C. Ng, G. Schussman - SLAC) --
  Future accelerators such as the NLC are designed to operate at high gradient to maximize efficiency and thus, reduce machine cost. At sufficiently high level, the RF power that provides the accelerating gradient in the linac also causes particle emission off the accelerator cavity wall. Subject to the same accelerating field, these surface-emitted electrons, called primaries, are either captured and accelerated down the linac, or returned to the cavity wall to generate secondary electrons. This cascading process produces a dark current (a parasitic beam consisting of both primaries and secondaries,) which can affect the main beam and increase the background signal in the detector down stream at the interaction point. In addition, excessively high surface fields also lead to RF breakdown which is damaging to the accelerator structure. Presently dark current and RF breakdown [19] are the limiting factors to high gradient in Linear Collider structure R&D, therefore theoretical studies which can shed light on these problems are of great importance.
• TraFiC4 (Andreas Kabel - SLAC) --
  Originated in DESY and under continuous development at SLAC, TraFiC4 [21] is a parallel 3D self-consistent code useful for optimizing the performance of Free Electron Lasers such as the Linac Coherent Light Source (LCLS) which is scheduled for construction at SLAC next year. These high current accelerators require very low transverse emittance of the incoming electron bunch which makes them highly susceptible to performance degradation by CSR effects that occur in the magnetic bunch compression sections. TraFiC4 calculates the CSR effects from first principles by using the retarded fields of weighted particles, and provides the high longitudinal resolution needed for handling large compression factor and eventually treating micro-bunching instabilities.