LAMMPS Stable Release 3 March 2020

Below is a list of major changes since the last stable release 7 August 2019.

General changes:

• The documentation file format has been changed from the custom txt2html markup to using reStructuredText (rst). This provides several advantages to have a better looking and more functional documentation. As a first step, all externally rendered equations have been converted to embedded math and all raw HTML markup converted. Most boxes with examples, will use syntax highlighting. Further improvements of typesetting and layout are forthcoming. Additional consistency checks are now applied when processing the documentation.
• Many file read operations are now monitored for read errors and thus allowing to detect damaged or incomplete files like potential parameter files and restart files
• Updates and improvements to the CMake build system to more closely match all the functionality provided by the conventional make based build system
• Support for threaded FFTs from FFTW and MKL, support for cuFFT and KISSFFT on the GPU with KOKKOS
• Expose neighbor lists to the library interface and the Python wrapper
• Documentation of installing LAMMPS via conda
• Many small improvements or bugfixes

Updates and new commands or styles

• new dump_modify options to embed ITEM: UNITS and ITEM: TIME with information about units and accumulated simulation time
• new pair styles cosine/squared, local/density, mesocnt
• new fix styles wall/reflect/stochastic, propel/self, 'npt/cauchy`
• new compute gyration/shape/chunk
• new compute hma for fast, high-precision computation of certain thermodynamic properties of solids
• new compute style centroid/stress/atom for computing per atom stress in a way allowing for more accurate heat flux computations with interactions involving more than two atoms
• new third_order command, a companion command to dynamical_matrix, for computing the third order force tensor from finite differences
• new options ratio and subset for create_atoms and similarly new options type/ratio and type/subset to the set command
• new tool for post-processing parallel tempering trajectories
• multiple updates, bugfixes, and improvements to fix bond/react
• improved implementation of minimization style fire
• support to run NWChem along with LAMMPS in client/server mode

Updates for packages

• KOKKOS now supports minimization (limited to minimization styles cg and quadratic), updates to the KOKKOS library, general improvements and bugfixes, more styles ported to KOKKOS
• several new SNAP potentials
• tweaks to USER-INTEL to compile correctly with too aggressively optimizing compilers, corrections and improvements
• bugfixes and improvements for the GRANULAR package
• several new SNAP potentials and a new compute snap for training SNAP potentials
• bugfixes and improvements to the KIM package, new command kim_param for accessing KIM model parameters
• updates to the USER-COLVARS package with updates to the included Lepton library
• updates to the USER-PLUMED package with bugfixes and support for recent releases
• updates to the USER-CGDNA package including a new coarse grained RNA model (oxRNA2)
• improvements and updates to the SPIN package, L-BFGS minimizer added
• updates, extensions, and improvements to the USER-MEAMC package
• memory leaks fixed when using styles from the GPU with multiple run commands, new pair style lj/cut/tip4p/long/gpu
• updates to fix rigid variants to add gravity to rigid objects with overlapping (granular) particles
• support for rerun and read_dump with USER-ADIOS

Backward compatibility notices

• building the USER-COLVARS package with all features enabled now requires a C++-11 compatible compiler
• due to changes in the random number generator initialization, results when using fix pour, fix deposit, fix evaporate or create_atoms random will not match with previous versions of LAMMPS. see PR #1569
• the gjf keyword for fix langevin no longer supports the option yes. Instead you have to use either vhalf or vfull
• the FIRE minimizer has been replaced with an improved implementation. The previous implementation is available as fire/old.
• new documentation has to be written in reStructuredText format