SODA basics

SODA provides a common computational foundation for independently developed scientific workflows. It supplies parallel task execution, resource coordination, and data collection. These pages explain how to use that infrastructure to run workflows such as APE, EZGA, NEBScape, and AKMC.

Each workflow defines its scientific methods and exploration strategy, with its own research contributions and references. SODA provides the shared infrastructure that supports their execution.

Start with the installation guide if you need a SODA build.

Analyze an existing database

If you already have a results database and want to inspect it, go directly to Output analysis.

How a calculation fits together

A workflow decides which calculations to perform and how their results contribute to the exploration. It creates tasks, such as relaxing a candidate structure or searching for a transition state. As results arrive, the workflow can select more tasks.

A potential or trained model describes the atomic interactions. A calculator provides the energies and forces used by the calculation. For example, an EAM potential can be evaluated through LAMMPS, while a supported machine-learning model may be evaluated through a LAMMPS or ASE interface. The workflow and the calculator are separate choices, with compatibility determined by the workflow integration and the installed build.

SODA distributes the tasks defined by the selected workflow among workers using MPI and collects their results in HDF5. The workflow supplies the scientific logic; the framework provides the execution and data-management infrastructure.

Term	Meaning in these tutorials
Run	One prepared calculation, with its own inputs, working directory, and outputs.
Workflow	The method that selects tasks and interprets their results.
Task	A unit of simulation work assigned to a worker.
Worker	An execution resource that performs tasks during the run.
Scheduler job	The CPU/GPU allocation requested from an HPC scheduler such as Slurm.

One scheduler job can execute many workflow tasks through SODA. The [Main] job setting in config.ini names the workflow; it does not describe the scheduler allocation.

Choose the scientific workflow

The following methods and frameworks can be run through the available SODA integrations. Their tutorial pages link to the original scientific work and, where applicable, the independently maintained software projects.

Workflow	What it explores	Starting point
APE	Elementary processes and connected states, guided by local atomic environments.	APE tutorial
EZGA	Candidate structures through evolutionary populations and generations.	EZGA tutorial
NEBScape	Surface-reaction structures, pathways, and transition-state barriers.	NEBScape tutorial
AKMC	Kinetic evolution through transitions between states; the implementation available through SODA is preliminary and used primarily for benchmarking and framework testing.	AKMC tutorial

Choose a potential appropriate for the elements and environments you want to study. Then use the installation selector to identify the required software. Installing a calculator does not establish that its model is suitable for a particular scientific system.

The SODA guides describe the supported EZGA and NEBScape integration paths. When adapting inputs from their native frameworks, use the conversion and configuration guidance linked from each workflow tutorial.

Follow a run from inputs to results

Page	What you will learn
Preparing a run	Select a build, create a run directory, configure the workflow, and check its inputs.
Parallel execution	Understand master/head/worker roles and choose a CPU/GPU launch layout.
Monitoring and outputs	Follow progress, locate results, and open a database.
Restarting a run	Resume a calculation from saved results and understand its recovery point.

For the infrastructure design, see the SODA manuscript. For the scientific methods, see each workflow’s original publications. Both are listed in Citing SODA.

Continue with Preparing a run.