Changes between version 1.3 and version 1.4.0

This is the list of changes made after the release of ASPECT version 1.3 for version 1.4.0. All entries are signed with the names of the author.

1. New: There is now an initial temperature field for the ellipsoidal chunk geometry model that allows discretizing the model domain into two layers separated by a user-defined isothermal boundary.
   (Tahiry Rajaonarison and D. Sarah Stamps, 2016/05/09)

2. New: The variable $ASPECT_SOURCE_DIR can now be used inside .prm files in various locations (including "include" statements).
   (Timo Heister, 2016/05/02)

3. Improved: Mesh refinement plugins can now update member variables at the beginning of each time step. This can be used to make things like time-dependent mesh refinement criteria, or fancy models that depend on calling an external program. The most immediate added functionality is the ability to define time-dependent minimum/maximum refinement functions.
   (Jonathan Perry-Houts, 2016/04/13)

4. New: There is a new function Utilities::orthogonal_vectors() that allows computing one (in 2d) or two (in 3d) orthogonal vectors to a given input vector.
   (Wolfgang Bangerth, 2016/04/13)

5. Improved: The particle sorting algorithm performance was considerably improved. It now sorts neighbor cells by the distance between the particle and the face center that is between the old cell and the neighbor and then checks if the particle is in this list of neighbor cells. Only if it is not found there, a search over all local cells is performed.
   (Rene Gassmoeller, Wolfgang Bangerth, 2016/04/12)

6. New: There is now a postprocessor "point values" that allows evaluating the solution at a number of user-defined evaluation points.
   (Wolfgang Bangerth, 2016/03/24)

7. Fixed: Combining particles with initial adaptive refinement steps used to create a multiple of the selected number of particles. This is fixed now.
   (Rene Gassmoeller, 2016/03/18)

8. Changed/new: The system by which linear systems are assembled has been rewritten into a collection of functions that each implement certain subsets of terms. This makes it far more flexible to combine different models, different formulations, etc.
   (Wolfgang Bangerth, Timo Heister, Rene Gassmoeller, Juliane Dannberg, 2016/01/15-2016/03/08)

9. Improved: The Introspection class has now a new base class FEVariableCollection that allows flexible modification of the finite element variables involved in a computation, even inside a plugin.
   (Timo Heister, 2016/03/08)

10. Fixed: The uniform radial and uniform box particle generators now produce globally unique particle IDs.
    (Harsha Lokavarapu, Gerry Puckett, 2016/03/04)

11. Fixed: The 'Simpler' material model produced floating point exceptions in models with compositional fields. This is fixed now.
    (Lev Karatun, Rene Gassmoeller, 2016/02/26)

12. New: The advection systems (for temperature and compositions) can now be discretized using the symmetric interior penalty discontinuous Galerkin method. This can be useful to explore solution without adding artificial smoothing. This is controlled by two new input parameters in 'Discretization': use_discontinuous_temperature_discretization and use_discontinuous_composition_discretization.
    (Sam Cox, 2016/02/22)

13. Changed: ASPECT by default wrote one output file per MPI process that was written in a background thread to a temporary location first and then moved to its final location. If any of the steps failed it tried again by writing directly to the output location. This approach needed complicated logic and did not succeed on all systems. In order to increase stability the new default behavior is to write straight to the output folder. This might decrease performance on clusters with slow network file systems. The old behavior can be recovered by setting 'Write in background thread' to true and set a temporary storage location by 'set Temporary output location'. Note that this functionality was and is only available if 'Number of grouped files' is set to its default value of 0, and therefore MPI-IO is not used for parallel output. For larger models with hundreds of parallel processes using MPI-IO is recommended.
    (Rene Gassmoeller, 2016/02/14)

14. New: Added 'command' postprocessor for executing arbitrary commands.
    (Jonathan Perry-Houts, 2016/02/11)

15. Improved: The option to increase the output resolution by linear interpolation of the quadratic elements now correctly uses the mapping of curved geometries to interpolate cells. This increases output accuracy for models that use curved geometries and use 'Set Interpolate output = true'. The simulation itself is not affected.
    (Rene Gassmoeller, 2016/02/08)

16. Changed: The GPlates plugin is restructured in the style of the AsciiData Plugin. The major difference is that the interpolation is now performed in spherical coordinates instead of Cartesian coordinates. Note that some input parameters have changed: "Time step" is now called "Data file time step", "Velocity file start time" is now called "First data file model time", "Interpolation width" does not exist any more, but there are three new parameters called "First data file number", "Decreasing file order" and "Lithosphere thickness".
    (Eva Bredow, Rene Gassmoeller, 2016/02/04)

17. New: ASPECT no longer relies on the availability of a command- processor (terminal) at run-time, by providing fallbacks to C commands. This adds support for architectures that do not offer a terminal on compute nodes (like IBM BlueGene/Q).
    (Rene Gassmoeller, 2016/02/03)

18. Changed: The 'depth' function of the 'box' geometry model and the 'two merged boxes' geometry model previously threw an exception when asked for the depth of a point outside of the initial model domain. This is not longer appropriate for models with free surfaces and therefore the behavior was changed to the behaviour of the 'spherical shell' geometry model, which is a cutoff of the depth to the range (0,maximal_depth).
    (Rene Gassmoeller, Sascha Brune, 2016/01/11)

19. New: There is now a parameter called 'Additional tangential mesh velocity boundary indicators' that allows to specify boundaries which elements are allowed to deform tangential to the boundary. This can be useful in models with free surface and a prescribed material in-/outflow at the sides. Previously in this case the uppermost element became distorted over time, now the whole boundary mesh adjusts according to the deformation. This change also fixes the handling of traction boundary conditions in models with free surface.
    (Anne Glerum, Rene Gassmoeller, Ian Rose, 2016/01/11)

20. Changed: The interfaces of the boundary composition and boundary temperature plugins have been deprecated. Their replacements not longer contain references to the geometry model, which was a leftover from an earlier development stage. Users should derive their plugins from SimulatorAccess if they need access to the geometry model. The deprecated functions will be removed in a future ASPECT release.
    (Rene Gassmoeller, 2016/01/04)

21. New: A new mesh refinement plugin was added that refines cells according to the density of particles in that cell.
    (Rene Gassmoeller, 2015/12/19)

22. Changed: The boundary_velocity(const Point<dim> &position) const function has now been deprecated in favor of the new function boundary_velocity (const types::boundary_id boundary_indicator, const Point<dim> &position) const.
    (Menno Fraters, 2015/12/16)

23. New: Visualization postprocessors for thermal conductivity and thermal diffusivity.
    (Anne Glerum, 2015/12/03)

24. New: The tracer architecture has been completely overhauled. It is now more flexible and allows for easier modification. Additionally tracers now carry properties with them, which allows for a variety of new use cases. A number of bugs related to curved cells in spherical models with tracers have been resolved.
    (Rene Gassmöller, 2015/12/01)

25. Fixed: Whenever the base models used by either the "depth dependent" or "averaging" material models depended on anything that requires accessing the simulator, then this led to segmentation faults. This is now fixed.
    (Wolfgang Bangerth, Shangxin Liu, 2015/11/24)

26. New: The tolerance of the preconditioners of the A and S block are now available as parameters in the prm file. There is now also a section added to the manual on how to use these parameters to make ASPECT in certain situation faster.
    (Menno Fraters, 2015/11/08)

27. Changed: The DynamicTopography postprocessor and visualization plugins now use the more accurate Gaussian quadrature rule for evaluating cell averages of the surface stress.
    (Ian Rose, 2015/11/04)

28. New: Add depth postprocessor which visually outputs the depth for all points inside the domain, as determined by the geometry model.
    (Menno Fraters, 2015/10/15)

29. New: The "Stokes residual" postprocessor will output the convergence of the Stokes residual of each linear solve.
    (Timo Heister, 2015/10/21)

30. New: The DepthAverage postprocessor can now output as a plain ascii text table.
    (Ian Rose, 2015/10/05)

31. Fixed: Free surface computations now work with checkpointing.
    (Ian Rose, 2015/09/29)

32. Fixed: certain combinations of boundary conditions with a free surface could result in accessing nonexistent matrix entries. This is fixed.
    (Ian Rose, 2015/09/28)

33. New: Ability to automatically resume models from a checkpoint if previous checkpoint files exist (Resume computation = auto).
    (Jonathan Perry-Houts, 2015/09/25)

34. Changed: the user can now select a subset of the laterally- averaged quantities to be computed in the DepthAverage postprocessor.
    (Ian Rose, 2015/09/04)

35. New: The history of the Stokes solver residuals is saved and can be accessed using the post_stokes_solver signal and will be written to a file automatically in case the solver doesn't converge.
    (Timo Heister, 2015/09/02)

36. Fixed: The laterally averaged sinking velocity and velocity magnitude calculations did not check whether the user selected m/s or m/yr for output values. Now they do.
    (Ian Rose, 2015/08/28)

37. Fixed: The lateral averaging of the velocity magnitude was mistakenly calculating the square of the velocity. Now it calculates the magnitude.
    (Ian Rose, 2015/08/28)

38. Changed: The interface of material models no longer declares property_depends_on() functions. The dependencies of parameters on solution variables are instead handled by a structure in the base class. All included material models have been updated. User written material models will continue to work as before. Since there is no solver yet that utilizes the other dependencies, the impact of this change is limited.
    (Kimberly Moore, Rene Gassmoeller, Wolfgang Bangerth, 2015/08/26)

39. New: The DepthAverage postprocessor now can calculate the laterally averaged heat flux in the interior of the simulation.
    (Ian Rose, 2015/08/24)

40. New: There is now a new initial condition in which the temperature field is perturbed following the SAVANI shear wave velocity model by Auer et al., 2014. The data were downloaded from http://n.ethz.ch/~auerl/research.html .
    (Shangxin Liu, 2015/08/20)

41. New: A box Geometry Model plugin with additional boundary indicators for the upper part of the box and corresponding Boundary Temperature and Composition Model plugins. With this plugin, different boundary conditions can be prescribed on the upper and lower part of the vertical domain boundaries.
    (Anne Glerum, 2015/08/14)

42. New: Plugin for visualizing the boundary indicators used by the Geometry Model.
    (Anne Glerum, 2015/08/14)

43. New: There is a new visualization postprocessor which displays the heat flux in the vertical direction, where upwards heat flux is positive.
    (Ian Rose, 2015/08/12)

44. New: A new material averaging option using logarithms is added. This is combined with the existing averaging schemes. Taking the viscosity for example, the log averaging will average 10^23 and 10^21 to 10^22.
    (Shangxin Liu, 2015/08/09)

45. New: A material model plugin for Drucker-Prager plasticity.
    (Anne Glerum, 2015/08/03)

46. Fixed: Quasi-implicit stabilization of a free surface had used the reference density instead of the density as evaluated by the material model. Now it uses the actual density of at the surface. This should not change much unless the density at the surface is significantly different from the reference density.
    (Ian Rose, 2015/07/22)

47. New: Plugin for visualizing groups of compositional fields as vectors.
    (Jonathan Perry-Houts, 2015/07/12)

48. New: For free surface computations there is an option to advect the mesh vertically (in the direction of gravity), in addition to the old formulation which advects it in the direction normal to the surface. This can be enabled by setting "Surface velocity projection" to "vertical" in the "Free surface" section of a parameter file. This scheme can maintain better mesh regularity properties for computations where there is a large deformation, or large curvature.
    (Ian Rose, 2015/07/10)

49. New: There is now an option in the Visualization postprocessor for outputting the mesh velocity in free surface runs.
    (Ian Rose, 2015/06/16)

50. New: There are now parameter files and a section in the manual for reproducing the benchmarks for free surface computations from Crameri et al. (2012).
    (Ian Rose, 2015/06/14)

51. New: One can now prescribe the traction on a boundary instead of supplying velocity boundary conditions. This is done in a similar way as for the prescribed velocity boundary conditions: For a given boundary indicator, one can prescribe all or a selection of the traction components.
    (Wolfgang Bangerth, Anne Glerum, 2015/05/29)

52. Changed: We now use the exact formulation with the compressible strain rate instead of an approximation using the right hand side of the mass conservation equation to calculate the shear heating. This is more accurate in compressible models.
    (Juliane Dannberg, 2015/05/29)

53. New: There is now a new geometry model called chunk, which takes radius, longitude (and latitude) pairs and creates a regional chunk of a spherical shell. Spherical boundary and initial conditions have been updated to accept this new model. The conversion conventions between [longitude, latitude and radius], [phi, theta and radius] and Cartesian [x, y, z] are consistent with mathematical convention and other models in dealii/ASPECT. This model was based largely on work on deal.ii by D. Sarah Stamps, Wolfgang Bangerth, and in ASPECT by Menno Fraters.
    (Bob Myhill, 2015/05/29)

54. New: Added the ability to prescribe internal velocities with an ascii file.
    (Scott Tarlow, 2015/05/29)

55. New: Three new material averaging schemes are added. These are combined with the existing averaging schemes, except for the Q1 averaging schemes into a compositing material model. The new averaging schemes are normalized weighted distance versions of the arithmetic, harmonic and geometric averages.
    (Menno Fraters, 2015/05/28)

56. New: There are now postprocessors BoundaryDensities and BoundaryPressures which calculate laterally averaged densities and pressures at the top and bottom of the domain.
    (Ian Rose, 2015/05/28)

57. New: Postprocessor and visualization postprocessor plugins can now state that they require other postprocessors to run as well, for example because they want to query information computed by these other postprocessors in computing their own information. This is done using the aspect::Postprocess::Interface::requires_other_postprocessors() function.
    (Wolfgang Bangerth, 2015/05/28)

58. New: Added cookbook to prescribe initial condition from shear wave velocity model.
    (Jacqueline Austermann, 2015/05/28)

59. Changed: The heating models have a new structure now: Instead of the implementation in the assembly, there is a heating plugin for each model now that can be used both in the assembly and the postprocessors, and a new heating model manager that combines the plugins by adding the individual heating terms.
    (Juliane Dannberg, 2015/05/27)

60. Fixed: There was a bug in the make pressure rhs compatibility function that caused the linear solver to fail in models with a significant in- or outflux of material. This is fixed now.
    (Juliane Dannberg, 2015/05/27)

61. New: Added cookbook for prescribed internal velocity values.
    (Jonathan Perry-Houts, 2015/05/26)

62. New: There is now a Material model called diffusion dislocation that implements diffusion and dislocation creep viscosity.
    (Bob Myhill, 2015/05/26)

63. Changed: Modified S40RTS initial condition file to incorporate the option to zero out heterogeneities within a given depth.
    (Jacqueline Austermann, 2015/05/26)

64. New: There is now a refinement plugin based on strain rate which will come handy to capture shear bands when combined with plasticity.
    (Cedric Thieulot, 2015/05/26)

65. New: There is now a Material model called Depth dependent that implements depth-dependent viscosity through the modification of any other Material model.
    (Jacqueline Austermann and Max Rudolph, 2015/05/24)

66. The manual.pdf is no longer part of the git repository but you can find it online at http://aspect.dealii.org or you can build it yourself.
    (Timo Heister, 2015/05/23)

67. New: There are now a set of global constants defined for physical properties and for radius and gravity parameters relevant to Earth and Mars.
    (Bob Myhill, 2015/05/22)

68. New: There is now a SimulatorAccess::get_statistics_object() function that allows all users of this class to record information in the statistics file, whether they are postprocessors or not.
    (Wolfgang Bangerth, 2015/05/22)

69. New: ASPECT now also provides a signals mechanism to attach user-defined functions to certain events over the course of a simulation. This allows more fine-grained observation and intervention to user additions. A new section in the manual explains how to extend ASPECT this way.
    (Wolfgang Bangerth, 2015/05/21)

70. Changed: The free surface handler now detaches internal manifolds for cases when the domain has them, since they are not necessarily a good description of the geometry when there has been large mesh deformation.
    (Ian Rose, 2015/05/21)

71. Changed: The documentation for nullspace removal is now more descriptive of what ASPECT is actually doing.
    (Ian Rose, 2015/05/21)

72. New: There is now a mesh refinement criterion called Slope, intended for use with a free surface, which refines where the topography has a large slope.
    (Ian Rose, 2015/05/21)

73. Changed: The specific heating plugin has a new interface now; it gets the material model inputs and outputs and fills a vector with heating model outputs for the whole cell.
    (Juliane Dannberg, 2015/05/20)

74. New: There is now an (3D) ellipsoidal chunk geometry model where two of the axis have the same length. The ellipsoidal chunk can be non-coordinate parallel part of the ellipsoid. This plugin is a joined effort of Menno Fraters, D Sarah Stamps and Wolfgang Bangerth
    (Menno Fraters, 2015/08/28)