Meep release notes
|License and Copyright|
Here, we describe what has changed between releases of the Meep package. You can also refer to the
NEWS file in the Meep package (or the
ChangeLog file for a more detailed listing).
20 July 2012
- Fixed to work with Guile version 2.x (older versions still work); requires libctl 3.2 or later.
epsilon-input-filefeature to read a scalar dielectric function from an HDF5 file (similar to MPB).
- Support for anisotropic dispersive materials (tensor σ parameter).
- Support for Drude dispersion model. New syntax is
make lorentzian-susceptibility, etc. (old
make polarizabilityis still supported for backwards compatibility).
- Support for "thermal" dispersive materials which include noise term in the polarization.
dft-ldosfeature for efficient LDOS-spectrum computation.
- Documented stress-tensor (force) spectrum computation feature.
mean-stretchproperty of PML (defaults to 1), to support real coordinate stretching for damping evanescent modes.
- Support for eigenmode-source feature using upcoming MPB release.
- Various small bugfixes.
24 August 2009
- Fixed release bug preventing Casimir calculation from running.
20 August 2009
- Meep's PML is now a true PML for arbitrary anisotropic, dispersive, and conducting media. (Now uses a slightly unconventional reformulation of PML described at ab-initio.mit.edu/meep/pml-meep.pdf)
- Fixed bug which caused anisotropic non-diagonal μ to be unstable.
- Fix compilation failure with gcc 4.4 due to missing
cstdioheader (thanks to Linran Fan and Bin Shao for the bug reports).
- C++ interface:
volumewas renamed to
geometric_volumewas renamed to
volume, to better reflect their respective roles.
accurate-fields-near-cylorigin?option to have more accurate fields near the r=0 origin for large m in cylindrical coordinates, at the expense of requiring a smaller Courant factor. (Default is
false, corresponding to behavior in older Meep versions.)
- In 2d computational cells, added much more efficient support for eikz z-dependence, enabled by new
special-kz?input variable (default is
falsesince it only works in 2d and is a little subtle for real fields).
- Includes preliminary new features to aid in computation of optical forces (both classical and quantum Casimir forces); further documentation pending more testing.
- Removed obsolete
docdirectory (all documentation is on the website these days).
- Small performance improvements in Lorentzian dispersion handling.
configurescript failure when cross-compiling.
- Fix compilation failure with MPICH.
5 June 2009
GUILE_CONFIGenvironment variable to override location of
configurescript; this is useful when cross-compiling.
2 June 2009
- Correct superficial
make checkfailure on 32-bit x86 machines with gcc 4.3.x, due to slight impact on floating-point rounding by automatic SSE/SSE2 vectorization; thanks to Silviu Popescu for the bug report.
- Correct superficial
make checkfailure when compiling under icc.
28 May 2009
- Enable correct operation and passed test suite when
MEEP_SINGLE(single-precision) mode is enabled in
meep.hpp; thanks to Seyoon Kim for the bug reports.
- Use new automake features to have less-verbose build output by default (you can build in verbose mode by
make V=1), and running all test programs then reporting which ones failed instead of stopping at the first failure.
- Fix superficial failure in
2D_convergencetest under gcc 3.4.6; thanks to Alex Prengel for the bug report.
- Fix failure in
fluxtest under gcc 4.3.1 in some cases; thanks to Alex Prengel for the bug report.
- Fix compilation problem with gcc 4.4, correcting Debian bug #505002.
28 March 2009
- New timestepping scheme for off-diagonal anisotropic epsilon and mu, based on technique by Werner and Cary [ J. Comp. Phys. 226, 1085 (2007) ], that improves FDTD stability when anisotropy is present (such as when subpixel averaging is used on isotropic media).
- Scheme user interface now supports user-specified anisotropic (real-symmetric) epsilon and mu (via
mu-offdiagparameters, similar to MPB). Accurate subpixel averaging of anisotropic media based on the method by Kottke, Farjadpour, & Johnson [ Phys. Rev. E. 77, 036611 (2008) ].
- Anisotropic dispersive materials are now supported, although currently the dispersive part of the epsilon/mu tensor must be diagonal, via the new
sigma-diagparameter of polarizability. (The corresponding C++ interface has also removed
delta-epsilonparameter of polarizability has been removed; you should use
fields::integrate2function (and corresponding Scheme function
integrate2-field-function) to perform integrations involving two simulations with the same computational cell (e.g. field-overlap calculations for coupled-mode theory).
- In the Scheme interface, subpixel averaging is not used for user-specified material-function types; you only get subpixel averaging for the standard shapes (blocks, cylinders, etcetera).
- Haskell code-generation is no longer used, and
hsrcdirectory is removed. Bitrotted and undocumented (hence unused) saturable-absorber feature has been removed, along with
energy-saturationparameter of polarizability.
- Some bug-fixes to test programs that made them overly sensitive to roundoff errors and possibly fail depending on the compiler. (New
meep-round-timefunctions to round times to single-precision, useful for robust time comparisons.)
17 March 2009
- Bug fix in cylindrical code, which caused it to blow up in some circumstances for nonzero m.
- Bug fix: non-integrated sources with conductivity are now second-order accurate, thanks to Alejandro Rodriguez.
- Bug fix in writing strings with parallel HDF5, thanks to Zheng Li for the bug report.
- Check that PML parameters are sensible (e.g. that total PML thickness is no greater than cell thickness) to avoid common mistakes.
extra-materialsinput variable, so that you no longer have to use "dummy objects" to specify the existence of some materials when using
24 July 2008
- Fixed circular dependency in
Makefile, which caused problems with some versions of
make; thanks to Kaoru Narita for the bug report.
21 July 2008
- Fixed incompatibility with Guile 1.6.x or earlier; thanks to the bug report by Andreas Unger.
20 July 2008
- Improved handling of nested
- Bug fix: parallel builds (
make -j) should now work.
- Bug fix: pkg-config file was incorrectly installed for MPI version; thanks to Majid Sodagar for the bug report.
19 July 2008
- Support for user-specified permeability (mu). Renamed "
dielectric" to "
medium" in libctl interface, new "
mu" property and new
output-mufunctions, and new "
Permeability" and "
Bx" etc. field types.
- Support for user-specified electric and/or magnetic conductivities. These are especially useful to add a desired dissipation loss (an imaginary part of ε/μ) in a narrow bandwidth, without messing around with Lorentzian dispersive materials.
- Add predefined
perfect-magnetic-conductor(μ = −∞) material, along with
perfect-electric-conductor(ε = −∞).
synchronized-magneticstep function to allow step functions to run with the electric and magnetic fields synchronized in time to second-order accuracy. See Synchronizing the magnetic and electric fields.
- New PML implementation (UPML instead of split-field), should have lower reflection in many cases.
- User-specified PML profile and asymptotic reflection.
- Internally, all timestepping code is now handwritten (and much shorter) rather than old verbose Haskell-generated code; this should make it easier to add new features.
- Add support for non-integrated current sources, if the
is-integrated?property of the current is set to false; this is now the default, to make handling of E and H sources more similar and intuitive.
- Work with HDF5 1.8 (which previously would not compile unless you manually set a preprocessor flag, due to API changes).
- Check for
/usr/include/ctl/ctl.h(default in Fedora), and check for libctl in
/usr/share/libctl3(default in Debian & Ubuntu).
- Bug fix: fixed relative phase of E and H sources (which were off from one another by half a timestep); thanks to M. Megens for bug report.
- Bug fix: make sure h5 filenames have unique timestep for cases where Δt is very small or very large.
13 November 2007
- Bug fix in
flux_in_box, which accidentally returned the flux multiplied by the number of processors, instead of the flux.
- Bug fix in epsilon averaging for structures including metals (ε < 0), fixing an instability.
- Bug fix in
output-pngwhen running in parallel (removing race condition).
- Fixed bug that disabled subpixel averaging for dimensions=1 (thanks to Mischa Megens for the bug report).
- Fixed bug that caused
output-tot-pwrto stop Meep with an error message; thanks to Vyacheslav Sokolov for the bug report.
at-everystep functions less susceptible to rounding errors; thanks to L. Le Guyader for the bug report.
- Fixed bug in dispersive media that wasted memory on parallel machines (the polarization memory was not parallelized); thanks to J. L. Silva for the bug report.
- Bug fix in
output-png+h5, thanks to a report by Chad Husko.
- Fixed several deadlocks that could occur when the parallel Meep is used with a serial HDF5 library (we continue to recommend using the parallel HDF5 library with parallel Meep, however). Thanks in part to Lingling Tang for his bug report.
- For maintainer-mode, improved detection of Haskell package names; thanks to Liang Huo for the bug report.
21 August 2006
eps-averaging?is now turned on by default (in libctl interface), using much-improved algorithm by Ardavan Farjadpour. This greatly improves accuracy, and also allows continuous tuning of geometric parameters. (See our upcoming paper in Optics Lett., with a preprint at Citing Meep.) New input variables
subpixel-maxevalto control the accuracy of the subpixel averaging.
- Support for χ(2) (Pockels) as well as χ(3) (Kerr) nonlinearities.
- Symmetries no longer require the cell size to be an even number of pixels. Previously, Meep exited with an error in this case, whereas now it simply adds an extra pixel to the cell size as needed.
with-prefixstep function to allow you to use a different
filename-prefixfor selected outputs.
- New feature for
output-png: built-in shell variable
$EPSthat refers to the last-output epsilon
.h5file, which you can use to easily add dielectric contours/overlays to the field output image.
output-png+h5function that outputs both
- New functions
field-energy-in-box(convenience wrappers around C++ functions).
- Bug fix in Kerr nonlinearity:
chi3was accidentally scaled by factor.
- Bug fix: if you specified three or more symmetries, at most two symmetries were used (ignoring the rest).
- Bug fix in
rotate2symmetry, which wasn't working correctly.
- Bug fix in
add-fluxfor multiple flux regions, thanks to K. Choi.
- Bug fix in
harminvwhere it wouldn't allow you to call
harminvmore than once for the same run loop; thanks to Aristos Karalis.
- Bug fix in
load-fluxthat prevented it from working properly without output directories, thanks to Karl Koch.
- Fixed abort that sometimes occurred due to rounding when the source was the same width as the cell (thanks to G. J. Parker).
- Fixed minor build problems on Cygwin, SGI, and other systems, thanks to Christopher Kang, Robyn Landers, Florencio Garcia, and others.
1 April 2006
- Initial public release.