femaXX - A parallel solver package for the real-symmetric Maxwell eigenvalue problem

V0.1

femaXX is a simulation tool for computing eigenmodes of large RF structures. It is the result of an ongoing collaboration of PSI and the Institute of Computational Sciences at ETH Zuerich. femaXX uses unstructured tetrahedral grids and Nedelec finite elements to discretise Maxwell's equations. The Jacobi-Davidson algorithm (JDSYM) is used for computing selected eigenpairs of the resulting large sparse symmetric eigenvalue problem. The LOBPCG algorithm is provided as an alternative to JDSYM.

femaXX is a parallel code for execution on distributed memory architectures and uses the MPI (Message Passing Interface) standard for communication between the processes, which is supported by all modern parallel computers.

Suggested reading

Multilevel preconditioned iterative eigensolvers for Maxwell eigenvalue problems
Peter Arbenz and Roman Geus. Institute of Computational Science (ETHZ) and Paul Scherrer Institute.
Appl. Numer. Math. 54 (2): 107-121 (2005).

On a Parallel Multilevel Preconditioned Maxwell Eigensolver
P. Arbenz, M. Becka, R. Geus, U. L. Hetmaniuk, and T. Mengotti. Institute of Computational Science (ETHZ), Paul Scherrer Institute and Sandia National Labs.
Technical Report 465, Institute of Computational Science, ETH Zürich, December 2004.
Available at http://www.inf.ethz.ch/research/disstechreps/techreports/.

Towards a Parallel Multilevel Preconditioned Maxwell Eigensolver
P. Arbenz, M. Becka, R. Geus and U. L. Hetmaniuk. Institute of Computational Science (ETHZ), Paul Scherrer Institute and Sandia National Labs.
To appear in the proceedings of PARA'04: Workshop on the State-of-the-Art in Scientific Computing. Lyngby, DK, June 20-23, 2004.

Multilevel preconditioners for solving eigenvalue problems occuring in the design of resonant cavities
Peter Arbenz and Roman Geus. Institute of Computational Science.
Technical Report 396, April 2003.
Download: PDF Document (219.7 Kbytes)

PyFemax: A Python Finite Element Maxwell Solver
Roman Geus, Peter Arbenz and Lukas Stingelin. Paul Scherrer Institute, CH-5232 Villigen.
PSI-Annual Report, VI:48-49, 2002.

The Jacobi-Davidson algorithm for solving large sparse symmetric eigenvalue problems with application to the design of accelerator cavities
Roman Geus. ETH Zurich.
2002. PhD Thesis No. 14734.
Download: PDF Document (4.3 Mbytes)

Solving Maxwell Eigenvalue Problems for Accelerating Cavities
Peter Arbenz and Roman Geus and Stefan Adam.
Phys. Rev. ST Accel. Beams, Volume 4, February 2001.
Available at http://prst-ab.aps.org/.

A Comparison of Solvers for Large Eigenvalue Problems Originating from Maxwell's Equations
Peter Arbenz and Roman Geus.
Numer. Lin. Alg. Appl.: 6(1):3-16, 1999.

Two-level hierarchical basis preconditioners for computing eigenfrequencies of cavity resonators with the finite element method
Peter Arbenz and Roman Geus.
Recent Advances in Numerical Methods and Applications II. Proceedings of the 4th International Conference on Numerical Methods and Applications ’98, pages 151–160, Singapore, 1999. World Scientific.

Eigenvalue solvers for electromagnetic fields in cavities
Peter Arbenz and Roman Geus.
High Performance Scientific and Engineering Computing, pages 363–373. Springer-Verlag, 1999. (Lecture Notes in Computational Science and Engineering, 8).

A Comparison of Solvers for Large Eigenvalue Problems Occuring in the Design of Resonant Cavities
P. Arbenz and R. Geus. Institut für Wissenschaftliches Rechnen, ETH Zürich.
Numerical Linear Algebra with Applications, 6, 3-16 (1999).

Parallel Solvers for Large Eigenvalue Problems Originating from Maxwell's Equations
P. Arbenz and R. Geus. Institut für Wissenschaftliches Rechnen, ETH Zürich.
Proceedings of the EuroPar '98, Springer-Verlag, Berlin, 1998. (Lecture Notes in Computer Science, 1470).
Download: PDF Document (338.3 Kbytes)

TR 275: Eigenvalue Solvers for Electromagnetic Fields in Cavities
St. Adam¹, P. Arbenz², and R. Geus².. ¹ Paul-Scherrer-Institut, CH-5232 Villigen. ² Institut für Wissenschaftliches Rechnen, ETH Zürich.
October 1997.
Keywords: cavities, eigenvalue problems, finite elements, Maxwell equations, particle accelerators, spourious modes
Pages: 33. Language: English.

Download: PDF Document (652.8 Kbytes)

A comparison of solvers for large eigenvalue problems originating from Maxwell’s equations.
Stefan Adam, Peter Arbenz, and Roman Geus.
High Performance Computing on Hewlett-Packard Systems (HiPer ’97), pages 235-244. Academic Computer Center Cyfronet-Krakow, 1997.

Directory structure

• src Main directory containing femaXX library source files.
• src/colarray 1D and 2D array C++ classes with BLAS wrappers.
• src/eigsolv Eigensolver code.
• src/fem Finite element code for Nedelec and Lagrange tetrahedral elements.
• src/matrixmarket Code for reading sparse matrice in MatrixMarket format.
• src/nr C++ implementations of selected Numerical Recipes algorithms.
• src/optparse Command line argument parsing code.
• src/pbe Timer code.
• src/tetmesh Code for managing the tetrahedral mesh.
• src/tut Unit test framework.
• src/utility Collection of utility routines.
• examples Collection of femaXX example and driver programs.
• test Unit tests.
• bin Utility scripts for genrating performance plots and testing femaXX calculations.
• doc Documentation directory.
• doc/api femaXX API documentation generated from source code using Doxygen.
• doc/install Installation guide.
• install_3pl Installation scripts for the third party libraries and tools.

Links to more information

A list of links to important portions of the femaXX code.

• femaxx_main.cpp contains instructions on how to use the femaxx_driver program.
• The FemaxxDriver class steers the whole computation and calls the corresponding substeps.

People

A list of persons who have contributed to the femaXX code:

• Roman Geus
• Martin Becka
• Dag Evensberget
• Ulrich Hetmaniuk
• Tiziano Mengotti
• Zenon Mathews

In addition source code by

• W. Evan Sheehan (optparse)
• Vladimir Dyuzhev (TUT)

has been incorporated into femaXX.

---