Add Rotational Initialization

As the title states.

July Updates

Minor updates, centered around IR prediction.

• New trajectories can be build with a dictionary instead of a .yaml file using build_from_dict().
• Properties can now be recorded from individual frames and written to output files. Currently this is only done with the molecular dipole.
• A few improvements to presto.external.
• Atoms can be deleted when building a trajectory from an older checkpoint file.
• XTB updates seem to have broken the test suite. I'm investigating further but for now (1) XTB tests are mostly disabled and (2) Travis CI is constantly panicking.

Hotfix – Type Issues

Unfortunately some numpy type issues snuck into the last release; now fixed.

Summer Updates - Barostat

Mainly behind-the-scenes bugfixes; the most major change is the addition of the Berendsen Barostat to permit NPT simulations.

Hotfix - QC Initialization

There were some variable name issues - so the previous release didn't work. Embarrassing.

June Updates

• Lots of backend tweaking and adding various parameters to the config file. (In process of documenting in CONFIG.md).
• More advanced quasiclassical initialization options.
• Building new trajectories now only requires presto.build().
• Changed license to GNU Public License 3.0.
• Added website.

April Updates Part II

The main update this release is the addition of a save_interval attribute to Trajectory, which permits one to save less frequently than every timestep. For instance, if timestep is 0.5 fs and save_interval is 100, then frames will be saved every 50 fs.

This is NOT a good idea if you want to study reaction coordinates, but should dramatically shrink the size of output files for equilibrations, where you really only care about processes on the timescale of the velocity autocorrelation function. Previously even a 100 ps equilibration on a medium-sized system was creating 3 GB checkpoint files, which were almost completely useless.

You can set this parameter as save_interval in the checkpoint file. You can also set forwards to False now.

(I also fixed a Controller bug which was messing up temperature gradients.)

Hotfix

There were a few minor but breaking errors in the last release, which have now been remedied. Sorry about that!

April Updates

Minor bugfixes and feature additions, mostly centered around improved error handling. Only a few new features:

• Replica exchange code is more functional now but still in pre-alpha.
• PairwisePolynomialConstraint now supports "min" and "max" interlocking-sphere biasing potentials.
• Quasiclassical initialization now checks for harmonicity and retries initialization if the expected and found energies are too different, as in PROGDYN/Jprogdyn.

Introducing presto 0.2.0

This release contains substantial refactoring, bug fixes, and new features, and accordingly moves us into 0.2.0.

Refactoring:

• Bash scripts are no longer needed for xtb/g16 integration. Instead, subprocess spawns and runs jobs accordingly in temporary directories created and destroyed automatically. Should lower disk usage and increase portability to new systems.
• The presto.config file options have changed to require specification of xtb/g16 executables.
• A new Controller class was created to supervise running of trajectories and simplify Trajectory classes.
• Trajectories now use fasteners.InterProcessLock for lockfiles.
• XTBCalculator objects using GFN-FF can store the autogenerated topology file and access it for future runs.
• Frames now store their elapsed time once run.

Bugfixes:

• Potential computation of forces and energy was incorrect and has been fixed. This was a major bug, and fixing it helps with energy conservation a lot.

New features:

• Added boilerplate for Check and Reporter classes to analyze trajectories on-the-fly (to be expanded upon later).
• Quasiclassical velocity initialization from transition states is now supported.