v1.3

Mainly a bugfix release

New features:

• New Python-based field_transform.py routine, which allows resuming of field transforms (which can take days for very large databases).
• New wavefield file format: "Surface" group is removed, since its variables were not used at all by Instaseis or MCKernel.
• Input file templates have correct values for Instaseis/MCKernel usage.
• Mainly bugfixes

Bugfixes

• Correct Q (80 instead of 143) in the lithosphere of internal PREM model
• Correction to the ak135f velocity structure below the 660.
• Write velocity models for MINEOS correctly.
• Write 2nd order discontinuities correctly into velocity model file.
• Fixes in Dipole stiffness terms

v1.1

First bugfix release of AxiSEM.

Changes to last version:

• Format of external (1D) models is changed and simplified
• Power-law Q in attenuation
• Full-field output for kernel wavefields and rdbm
• Better checking of user input
• Cleaning up output

Bugfixes

• Serial mode now functional and compiling without MPI
• Bugfixes in NetCDF output of (kerner-related) wavefields
• Bugfix of all nodes writing into the same file, receiver.dat
• Bugfix in writing out 1D model files
• Bugfix in mesh decomposition for fluid spheres
• Fixed SOLVER/submit.csh script for arbitrary DATA directories, especially for 'single' simulations

v1.0

First general release of AxiSEM.

Working:

• 1D models included (PREM, IASP91, AK135 with different modifications)
• External 1D models
• Models can include solid and fluid zones
• Axially symmetric heterogeneities
• Attenuation
• Anisotropy
• General moment tensor and force sources
• Receiver locations on the surface
• Benchmarked for elastic models against yspec (Al Attar & Woodhouse, 2008)
  up to 1 Hz
• Newmark time integration scheme tested and benchmarked against YSPEC
• Symplectic time integration scheme tested, but not fully benchmarked
• NetCDF output
• Postprocessing to produce ASCII seismogram files
• Mesh decomposition up to 32 processors radially and 500+ (depending on
  frequency) laterally

Known issues:

• Other time integration schemes are not reliably working.
• Source time function 'quheavi' is not working reliably.
• Receivers must be on surface.
• Uppermost layer of model must be solid.