The source term Q fed to the dispersion models is built in two steps: first the radionuclide inventory A in the core, then the fraction of that inventory reaching the atmosphere in an accident. RADOSE solves the first step as physics; the second is scenario-dependent and stays in the user's hands.

The nuclides fall into two behavioural families. Short-lived ones (Xe-133, I-131) settle into a production–decay equilibrium while the reactor runs: their amount depends only on thermal power, independent of operating history. Long-lived Cs-137 (T½ = 30.08 y) accumulates instead: its amount scales with the total energy produced (power × core age).

From the single-nuclide production–loss balance (the saturation limit of the Bateman equation) at constant power, for t ≫ T½ the activity equals the production rate:

Reaching saturation takes ~5·T½: about 26 days for Xe-133, 40 days for I-131. For any operation uninterrupted over the last ~40 days the coefficients are history-independent. They are not pure U-235 yields — in burnt fuel part of the fissions come from Pu-239; the effective yields carry that correction (see Calibration).

Cs-137 barely decays on reactor timescales; every atom produced accumulates. Its activity ties to the total energy produced via the total fission count:

Core age (EFPD, effective full-power days) is the batch-strategy-derived mean age of the EOC mixture: age = Σ(batch share × cycles resident) × cycle EFPD. This single number carries the entire Cs-137 difference between designs:

The effective yields are calibrated against four independent OpenMC depletion campaigns (ENDF/B-VII.1 chain; three different fuel-management schemes); the coefficient band across the four designs is ±1%. The +15% correction on I-131 is the largest Pu-239 fission-share effect, folded into the coefficient:

The equations are verified against the four calibration campaigns and by three independent tests (the VSMR-300 blind test; CAREM-25 and IRIS out-of-envelope real designs) across the full SMR operating-power band (100–1000 MWth ≈ ≤300 MWe); across 7 independent points the largest deviation is 1.5%. All result tables live in the Source Term tab of the Benchmark page; the in-app coefficients are pinned by the `check:inventory` regression.

The step from inventory to source term uses the five-factor formula. RADOSE deliberately leaves these fractions to the user: the validated physics (inventory) is built into the tool, while scenario assumptions (fractions) remain transparent inputs.

A guiding example: the PWR in-containment release fractions of NUREG-1465 / RG 1.183 (release into containment; containment leakage is entered separately via LPF). These are a literature reference, not a recommendation — adapt them to your scenario:

The validity envelope and known limits of the equations — outside these conditions the coefficients are an extrapolation and a real depletion calculation is required: