Iterative Simulations

Use an iterative simulation when several calculations share most of a model but change selected inputs between runs. Common applications include parameter sweeps, source updates, feedback iterations, and optimization studies.

The workflow reuses one mcdc.Simulation and its Python model objects. Each iteration still prepares a complete, consistent simulation for transport; MC/DC does not patch only the changed value into a previous execution.

Build the Shared Model Once

Construct the geometry, materials, sources, tallies, settings, and techniques as usual. Attach them to one simulation before beginning the iteration. Objects that do not change can remain attached throughout the study.

Use separate simulations only when the calculations require independent object graphs. See Simulation Lifecycle for the normal construction workflow and process-ownership rules.

Update, Compile, and Run

Change the relevant Python objects at the beginning of each iteration. For example, the relative probabilities of two sources can be varied while the geometry, material, and tally remain unchanged:

source_mixes = (
    ("left_20", 0.2),
    ("left_50", 0.5),
    ("left_80", 0.8),
)

for case_name, left_fraction in source_mixes:
    source_left.probability = left_fraction
    source_right.probability = 1.0 - left_fraction
    simulation.settings.output_name = f"source_mix_{case_name}"

    simulation.compile()
    simulation.run()

The explicit compile call establishes a fresh model snapshot after each partial update. It is particularly useful when the updated model will be inspected or visualized before execution.

simulation.run() automatically compiles an unprepared simulation and marks it unprepared again after execution. A simple loop that only updates and runs may therefore omit the explicit compile call. If an attached object is modified after an explicit compilation or visualization, compile again before inspecting, visualizing, or running the updated model.

Keep Outputs Distinct

Assign a unique simulation.settings.output_name before every run so that a later iteration does not replace an earlier result. Keep tally names stable when the same post-processing operation will be applied to every output.

The driver or post-processing script can record the iteration parameters next to each output name. This makes plots and comparisons reproducible without depending on internal object identifiers.

Complete Example

See Iterative Source Reweighting for a complete runnable input. Its post-processing script overlays the flux profiles from three source mixtures and checks their expected symmetry.