wavepacket 0.5

A package for solving Schrödinger and Liouville von Neumann equations.

Description

Wavepacket is a Python package to simulate small quantum systems. Technically, it allows you to numerically solve Schrödinger and Liouville-von-Neumann equations for distinguishable particles.

The full documentation can be found under https://wavepacket.readthedocs.io. See also the advanced use cases for examples of what can be done.

Wavepacket focuses on a particular niche:

• Direct solution of Schrödinger / Liouville von Neumann equations. No electronic structure, no MCTDH, no semiclassics, at least for now.

• Grid-based representation of operators (DVR approximation). Atomic units to simplify equations.

• Accessibility is an overriding goal. Wavepacket should be directly usable in teaching. We try to provide good tutorials / docs, but also goodies like simple plotting.

• Exotic use cases should be as frictionless as possible. For example, switching between wave functions and density operators is supported well. Or random thermal wave functions are not much more complex to use than Gaussian states.

• Performance is a subordinate goal. For specific systems, you can easily improve the performance by integer factors.

Links

Documentation:

https://wavepacket.readthedocs.io

Source:

https://github.com/ulflor/wavepacket

Issue tracker:

https://github.com/ulflor/wavepacket/issues

Support

There are three different avenues for feedback and support:

• The preferred option is through the mailing list [e-mail]. You may want to subscribe at https://mail.python.org/mailman3/lists/wavepacket.python.org/ to ensure you receive all answers.

• You can use the Forum on github: https://github.com/ulflor/wavepacket/discussions You need a github account to post messages.

• You can open a new issue for bug reports or feature requests directly at https://github.com/ulflor/wavepacket/issues. This requires a github account. You can of course also report bugs or request features on the mailing list or in the forum. Depending on the complexity of the feature, this will lead to an immediate, rapid, or prioritized implementation.

Contribution

I currently lack a formal procedure for new contributors, but you are very welcome to contribute to the project. If you do not know what to do, use one of the support channels from the previous section ; there is enough work for multiple developers, also for non-coding skills (there is never enough documentation, for example).

History

The original long-time version of Wavepacket was written in Matlab and is still actively maintained under https://sourceforge.net/p/wavepacket/matlab. It is stable, battle-tested and works.

However, Matlab is expensive, and not all users had (legal) access to it. Also, the project’s architecture did not support some advanced use cases well. Finally, C++11 had just come out and looked cool, so I started a reimplementation in C++ around 2013, adding Python bindings as an afterthought. The C++ project will be superseded by this Python package, but can be found under https://sourceforge.net/p/wavepacket/cpp.

This worked really well, except that deploying C++ code is difficult. In particular, there was no cheap route towards building a “good” Python package, especially under Windows, which restricted accessibility.

The solution is this Python-only version based on numpy. The Python version is slower by a factor of two to three compared to C++-backed code. This is, however, often cancelled by a parallelization of the tensor operations.