PySDM

What PySDM is

PySDM is a free software package for simulating particle populations as they grow through diffusion or collisions, or break apart. Its development centers on atmospheric cloud physics, including aerosol, cloud, and rain microphysics. It uses the Super-Droplet Method Monte Carlo algorithm to represent collisional growth and is designed as a building block for simulation systems involving dispersed phases.

The documented compute backends include multithreaded CPU processing, CUDA on a GPU, and JAX, which is early in development. The example gallery spans condensation, coagulation and breakup, drop freezing, isotopic fractionation, aqueous chemistry, and seeding. It also demonstrates box, parcel, single-column, and two-dimensional prescribed-flow setups. PyMPDATA handles advection in the single-column and two-dimensional prescribed-flow examples.

Install PySDM from PyPI with pip or use the documented local Git checkout route. Example notebooks can be launched on Google Colab or mybinder.org, though the notebooks are not bundled in the examples package wheels. The GPU backend requires NVIDIA hardware and a CUDA driver. PySDM and PySDM-examples are free/libre open-source software under GNU GPL v3.

Who it suits

PySDM suits researchers and developers building simulations of dispersed phases, especially those working on atmospheric cloud, aerosol, or rain microphysics. Its range of process and model examples can help users explore different setups, while the backends offer CPU, CUDA GPU, and early-stage JAX options.

Limits

The JAX backend is early in development. The CUDA GPU backend requires NVIDIA hardware and a CUDA driver; example notebooks are not included in the examples package wheels.

Getting started

1. Open the PySDM website.
2. Install the package from PyPI with pip, or follow the local Git checkout instructions.
3. For GPU computing, use NVIDIA hardware with a CUDA driver.
4. Launch example notebooks through their Google Colab or mybinder.org badges.

Verdict

Choose PySDM if you need an open-source simulation building block for dispersed phases, particularly atmospheric cloud physics, and value its process examples and compute backend options. Look elsewhere if you need a mature JAX backend or cannot meet the NVIDIA and CUDA requirements for GPU use.