Theory & Method
Xe-135 / Sm-149 Poisoning Simulator · AntraSIM-0070 · version 1.0.2 · build 2026-09-20. This document describes the model as implemented in the application. Where the source material (the Jupyter prototype, release 1.0.1) and the executable model disagree, §2 says which was implemented and why.
How to read this document. Each topic follows the progression Intuition → Physical mechanism → Mathematical formulation → Engineering interpretation → Assumptions and limitations. A newcomer can stop after the first two layers; an experienced engineer finds equations, units, numerical implementation, data provenance and limits in the rest. Tables with computed numbers (§5, §7, §8, §10) are generated at build time by the same core code that the application runs.
Model status. Principle simulator. The equations are verified numerically (§7: closed-form solutions and an independent adaptive integrator). The model is not validated against measurements from a reactor, and the one-group cross sections, flux scaling and fission cross section are assumptions (§8, §10). It is not safety-grade and must not be used for core management, safety analysis, licensing or decisions in a real plant.
1. Physical Background
Intuition Fission creates fission products, and two of them absorb thermal neutrons so strongly that a very small concentration acts as a noticeable negative reactivity. Xenon-135 changes on a timescale of hours and gives the familiar xenon peak after a power reduction; samarium-149 changes on a timescale of days, is stable, and stays in the core after shutdown. The simulator shows how the two inventories and their reactivity effect follow an arbitrary power history.
Physical mechanism Most Xe-135 does not come directly from fission but from β− decay of I-135 (T½ ≈ 6.58 h); Xe-135 (T½ ≈ 9.14 h) decays to Cs-135 or is destroyed by neutron capture to Xe-136. Sm-149 is fed by β− decay of Pm-149 (T½ ≈ 53.08 h) and is destroyed only by neutron capture to Sm-150.
effective mass-135 source→I-135→Xe-135→Cs-135 · or (n,γ) Xe-136
effective mass-149 source→Pm-149→Sm-149→(n,γ) Sm-150
Short-lived precursors upstream of I-135 and Pm-149 are not modelled explicitly; their contribution is included in effective yields (§8, §10). The causal chain is: power → reference flux → fission-rate density → precursor production and loss → poison concentration → neutron absorption → reactivity effect.
Engineering interpretation After a power reduction the capture loss of Xe-135 falls at once while stored I-135 keeps feeding it, so xenon rises for several hours and the poison worth becomes more negative. After a shutdown the Sm-149 inventory grows towards the sum of the initial Pm-149 and Sm-149 inventories and never decays. A restart near the xenon peak may therefore need substantial positive reactivity. Assumptions and limitations Zero-dimensional, one-group, fixed cross sections, no feedback (§9).
2. Assessment of the Source Models
Scope Two source models were read completely: the Jupyter prototype (Antranova, v1.0.0) and the previous release of this application (v1.0.1). Their constants and code were compared with evaluated data and with each other; the executable model of release 1.0.1 was kept and its numerics were reproduced bit for bit (§7).
Table 2.1 — Model constants: Jupyter prototype v1.0.0 versus the release-1.0.1/1.0.2 defaults (T½ and the ratio are computed from the constants).
| Symbol | Unit | Prototype v1.0.0 | Release default | Release / prototype | T½ prototype → release |
|---|
| yI | atoms/fission | 6.338e−2 | 6.28187e−2 | 0.991× | |
| yX | atoms/fission | 2.600e−3 | 2.56630e−3 | 0.987× | |
| yPm | atoms/fission | 1.500e−3 | 1.08163e−2 | 7.21× | |
| λI | s⁻¹ | 2.870e−5 | 2.92615e−5 | 1.02× | 6.709 h → 6.58 h |
| λX | s⁻¹ | 2.090e−5 | 2.10657e−5 | 1.01× | 9.212 h → 9.14 h |
| λPm | s⁻¹ | 1.380e−5 | 3.62737e−6 | 0.263× | 13.95 h → 53.08 h |
| σa,X | cm² | 2.600e−18 | 2.60000e−18 | 1× | |
| σa,Sm | cm² | 4.300e−20 | 4.10000e−20 | 0.953× | |
| Σf | cm⁻¹ | 1.697e−1 | 1.00000e−1 | 0.589× | |
| φ1 | n/(cm² s) | 2.411e11 | 3.00000e11 | 1.24× | |
| νf | – | 2.430e0 | 2.43000e0 | 1× | |
| βeff | – | 6.500e−3 | 6.50000e−3 | 1× | |
Table 2.2 — Audit findings.
| ID | Finding | Resolution |
|---|
| A-1 | The prototype uses yPm = 0.0015, a factor 7.2× below the cumulative mass-149 chain yield obtained from the evaluated independent yields (Table 10.1). Sm-149 production is therefore underestimated. | Release default 0.0108163 retained; the prototype value is not used anywhere. |
| A-2 | The prototype uses λPm = 1.38×10−5 s−1 (T½ ≈ 14 h); the evaluated half-life of Pm-149 is 53.08 h (λ = 3.627×10−6 s−1), a factor 3.8× lower. | Release default retained. |
| A-3 | The prototype uses different generic assumptions (Σf = 0.1697 cm−1, φ1 = 2.411×1011 n/(cm² s) per %, σa,Sm = 4.3×10−20 cm²) and yields (yI = 0.063377, yX = 0.0026). | Release-1.0.1 values retained as documented in §8; they are assumptions with a coherent power density (§8). |
| A-4 | Release 1.0.1 built its pop-out windows by streaming HTML into a new document (a pattern that antivirus heuristics flag), used a generic placeholder logo, had no in-app verification suite, no step-count limit and no upper-bound (instability) check, an unversioned configuration file without unit suffixes, and a URL-hash state that was applied without the all-or-nothing validation of a configuration file. | Pop-outs built with DOM APIs; canonical logo; verification suite T1–T13; step cap and physical bounds with explicit messages; IC schema 2 with an explicit, tested migration of schema 1; URL-hash state removed (the IC file replaces it). |
| A-5 | Release 1.0.1 exported only a results table. | CSV now carries traceability metadata, an event column and the complete IC as a comment line; the power profile has its own CSV format. |
3. Governing Equations
Intuition Each nuclide is produced by fission or by decay of its parent and is lost by decay or by neutron capture. Production and capture scale with the flux; decay does not.
Mathematical formulation The flux follows the power linearly,
where P(t) is power in % of rated and φ1 the flux per percent power [n·cm−2·s−1]. With Σf the macroscopic fission cross section [cm−1], y the effective yields [atoms/fission], λ the decay constants [s−1] and σa the microscopic absorption cross sections [cm2], the concentrations [atoms·cm−3] obey four coupled linear ordinary differential equations:
The term Σfφ is the fission rate density [fissions·cm−3·s−1]. The Xe-135 equation contains the effective non-I-135 (bypass) production plus the explicit decay of I-135; the effective Pm-149 yield includes omitted short-lived mass-149 precursors, so no separate direct Sm-149 source is added (Table 10.1). For piecewise-constant flux each equation has a closed-form solution; for I-135, for example,
and at zero flux the Sm-149 inventory rises to the sum of the initial Pm-149 and Sm-149 inventories:
Engineering interpretation Xe-135 has two loss channels of comparable size at rated power, so its equilibrium depends on the flux; the Sm-149 equilibrium is independent of the flux because both its source (through Pm-149) and its capture scale with it. Assumptions and limitations Linear in power, fixed one-group cross sections, no burnup dependence, no spatial redistribution.
4. Inputs and Operating Conditions
Profile semantics The active power profile is a list of rows: each row holds a power level for its hold time and is then joined to the next row by a linear ramp of the given duration; a zero ramp is a step, and the ramp of the last row is not used. Two definitions are always stored and only the active one drives the run: the simple profile (three levels with two ramps) and the table profile (1–200 rows, edited in a pop-out window with CSV import/export or by dragging the blue points on the power plot). Both are converted to one canonical sequence of holds, ramps and steps before the solver runs; equivalent definitions give identical results (T5), and a zero-magnitude input changes nothing (T6). Power is 0–200 % of rated, each duration 0–1000 h and the whole run at most 20000 h.
Initial state Equilibrium at P₀ means infinite operation at the first row's power before t = 0 (equations in §3 with all derivatives set to zero):
If the first power is zero no equilibrium exists and the run starts from a clean core; Clean core starts with all four concentrations at zero.
Units Times are stored in hours and can be shown in hours or days; power in % of rated (rated thermal power is used for the MWth axis and columns only and does not set the flux); worth in pcm and in dollars.
Input Configuration (IC) and CSV The IC is a JSON file (format “AntraSIM-input-configuration”, schemaVersion 2) with unit suffixes in every key and both profile definitions. Loading is all-or-nothing; files of schema 1 (release 1.0.1) are migrated by a registered migration, tested by T10, and their missing inspection time is set to the end of the run; other schemas are rejected with a message. Results CSV files contain metadata, an event column and the complete IC (§5).
5. Calculated Quantities
Poison worth A first-order one-group estimate: the macroscopic absorption of the poison divided by the neutron production from fission, with negative sign:
where βeff is the effective delayed-neutron fraction; one dollar is βeff, so the conversion is model dependent. The expression is useful for trends and orders of magnitude; it is not a perturbation-theory calculation and omits adjoint weighting, leakage and spatial or spectral effects [5,6]. Δρ is measured from the poison-free state.
Milestones The most negative Xe-135 worth (with time), the largest deficit of the total worth relative to its value at the start, and the values at start and end are taken over every RK4 step. For the default case (scram after 5 h at rated power) the largest xenon worth is −3977 pcm at 13.34 h; the largest deficit relative to the start is −1874 pcm (−2.88 $) at 13.39 h, and the total worth at the end of the 105 h run is −568 pcm. The application also reports concentrations relative to equilibrium at the reference power and the equilibrium worth for that power.
Table 5.1 — Preset scenarios (generated from the code with the default model values).
| Scenario | Run [h] | Min. Xe worth [pcm] at t [h] | Deficit vs. start [pcm] at t [h] | End [pcm] |
|---|
| Scram from 100 % | 105 | −3977 at 13.34 | −1874 at 13.39 | −568 |
| Load follow 100 → 50 → 100 % | 76 | −2692 at 10.47 | −579 at 10.52 | −2565 |
| Trip and restart after 8 h | 71 | −3975 at 10.01 | −1871 at 10.01 | −2565 |
| Startup from clean core | 162 | −2119 at 162 | −2256 at 162 | −2256 |
6. Numerical Method
Approach Classical fourth-order Runge–Kutta with a fixed step per segment. Every hold boundary, ramp endpoint and step is represented exactly; for a ramp the flux is evaluated at the RK4 stage times. The nominal step is 30 s and is reduced to Δt = 0.05/kmax when edited values increase the fastest rate kmax = max(λI, λX + σa,Xφmax, λPm, σa,Smφmax). For the default case Δt = 30 s, kmaxΔt = 0.00297 and 12600 steps are taken; about 0.75 million steps per second were measured in Node, so the model is recalculated after every input change.
Stability and safeguards Explicit RK4 is stable for kΔt up to about 2.79; the margin is 2.79/0.00297 ≈ 937 at default. A warning appears when kmaxΔt > 0.02. The only non-smooth operation is clipping of negative concentrations to zero; it is inactive in the default case (0 clips, T7) and any activation is counted and reported. A concentration exceeding its physical saturation bound, or a non-finite value, stops the run with a message (T13). The number of steps is capped at 4×10⁶; exceeding it refuses the run with a message before any computation.
7. Verification and Numerical Accuracy
Scope Verification shows that the equations are solved correctly and that reference values are reproduced; it is not validation. The suite runs in the application (footer → Verification) and in Node. A failing test is a finding: the cause is diagnosed and the criterion is never adjusted to fit.
Table 7.1 — Release verification suite (13 of 13 passed; generated from an actual run).
| ID | Test | Criterion | Result | Measured |
|---|
| T1 | Equilibrium at P₀ is preserved for 60 h of constant power (all four concentrations) | max relative drift ≤ 1e-9 | PASS | 0 |
| T2 | Piecewise-constant power (steps, clean and equilibrium start): closed-form solution with expm1-stable exponential integrals | max error / run maximum ≤ 1e-10 (every third stored point) | PASS | 1.93e-13 |
| T3 | Multi-segment profile with ramps and steps: RK4 (30 s) vs. independent adaptive Dormand–Prince 5(4), rtol 1e-12 | max relative difference of the final state ≤ 1e-9 | PASS | 1.15e-13 |
| T4 | Time-step sensitivity: Δt = 240, 120, 60 s (pre-asymptotic range, above the round-off floor) against the Dormand–Prince reference | observed order 3.5–4.5; error at 240 s ≤ 1e-7; default 30 s ≤ 1e-11 | PASS | p = 4.02, 4.04; e(240) = 2.34e-10 |
| T5 | Equivalent input modes: simple profile = table profile with the same rows (default and all four scenarios) | identical results, max relative difference ≤ 1e-12 | PASS | 0 |
| T6 | Zero-magnitude inputs change nothing: a zero-duration power spike between two rows; a ramp between equal levels equals a hold | max relative difference ≤ 1e-12 | PASS | 0 |
| T7 | Finite, non-negative concentrations, negative poison worth, and no clip activation in the four scenarios and a 200 % case | all stored points; clip activations = 0 | PASS | pass; clips 0 |
| T8 | Scale invariance: (φ₁, Σ_f, σ) → (kφ₁, Σ_f/k, σ/k) leaves all states and worths unchanged; the rated power only rescales MWth; doubling all yields doubles all concentrations | max relative difference ≤ 1e-10 | PASS | 4.04e-16 |
| T9 | Analytic identities: Sm-149 equilibrium worth = 10⁵·y_Pm/ν_f at any power; 1500 h at 100 % from a clean core approaches the closed-form equilibrium (I, Xe, Pm) | max relative difference ≤ 1e-6 | PASS | 3.11e-9 |
| T10 | IC round trip restores every input exactly; 8 malformed files are rejected; a release-1.0.1 file (schema 1) is migrated explicitly and reported; a foreign schema-1 file is rejected | exact; 8 of 8 rejected; migratedFrom = 1 | PASS | exact; rejected 8/8; migration exact |
| T11 | Power-profile CSV: exact round trip in h and d; ';' delimiter with decimal comma; header-less file; 8 malformed files rejected | exact; 8 of 8 rejected | PASS | ok |
| T12 | Default-case regression values (Xe-135, Sm-149, total worth at 0 … 105 h) | max relative difference ≤ 1e-9 to the stored values | PASS | 0 |
| T13 | Failure handling: a forced unstable time step (Δt·k = 3.0 > RK4 limit 2.79) stops with a diagnostic and stores nothing; an excessive step count is refused with a message | both refused with the expected message; no partial results | PASS | announced; refused |
Table 7.2 — Default-case regression values (scram scenario, stored in the suite and printed from the same run).
| t [h] | 0 | 2 | 5 | 6 | 10 | 25 | 50 | 105 |
|---|
| Xe-135 [atoms/cm³] | 1.9800e15 | 1.9800e15 | 1.9800e15 | 2.4551e15 | 3.5137e15 | 2.6834e15 | 6.2213e14 | 1.2092e13 |
| Sm-149 [atoms/cm³] | 2.6381e16 | 2.6381e16 | 2.6381e16 | 2.6497e16 | 2.6947e16 | 2.8437e16 | 3.0356e16 | 3.2903e16 |
| Δρ total [pcm] | −2563.69 | −2563.69 | −2563.69 | −3073.98 | −4214.22 | −3350.96 | −1177.84 | −568.09 |
Continuity with release 1.0.1 The release-1.0.1 model block was executed side by side with the present core for the four scenarios, a 200 % multi-row table profile and a case with edited constants: the number of points and the extremes were identical and the largest relative difference of any stored quantity was 0 (bit-identical) (6 cases). The numerics are therefore unchanged; the additions in this release are checks, messages and interface.
8. Model Parameters and Constants
Evidence classes: externally sourced derived assumption not verified. All values are editable; the table refers to the defaults.
| Parameter | Default [unit] | Class | Basis |
|---|
| yI effective cumulative I-135 yield | 0.0628187 atoms/fission | derived | ENDF/B-VIII.0 U-235 thermal cumulative yield [3], as adopted in 1.0.1; reproduced within 0.27 % from independent yields with Sn/Sb branching [4] (Table 10.1). |
| yX effective Xe-135 bypass yield (cumulative Xe − cumulative I) | 0.0025663 atoms/fission | derived | 0.065385 − 0.0628187; independent Xe-135 plus 99.7 % of Xe-135m gives 0.002561 [4] (Table 10.1). |
| yPm effective cumulative Pm-149-chain yield | 0.0108163 atoms/fission | derived | Cumulative mass-149 chain; independent Pm/Sm yields are negligible [4]; recomputed 0.010824 (Table 10.1). |
| λI I-135 decay constant | 2.9261532e−5 s⁻¹ | externally sourced | T½ = 6.58 h; ICRP-107 6.57 h [1], NuDat export 6.55 h [2] (Table 10.2); origin of 6.58 h not established. |
| λX Xe-135 decay constant | 2.1065742e−5 s⁻¹ | externally sourced | T½ = 9.14 h; equals ICRP-107 [1]; NuDat export 9.168 h [2]. |
| λPm Pm-149 decay constant | 3.6273716e−6 s⁻¹ | externally sourced | T½ = 53.08 h; identical in ICRP-107 [1] and NuDat export [2]. |
| σa,X Xe-135 effective absorption (depletion) | 2.60e−18 cm² | assumption | 2.6×10⁶ b, rounded thermal-scale one-group value; spectrum dependent; not checked against the evaluated library in this release. |
| σa,X,w Xe-135 effective absorption (worth) | 2.60e−18 cm² | assumption | Equal to the depletion value by default. |
| σa,Sm Sm-149 effective absorption | 4.10e−20 cm² | assumption | 4.1×10⁴ b, rounded thermal-scale one-group value; spectrum dependent. |
| Σf macroscopic fission cross section | 0.1 cm⁻¹ | assumption | Representative fuel region; with φ1 it fixes the power density (below). |
| νf neutrons per fission | 2.43 – | assumption | Generic LWR mix. |
| φ1 flux per percent power | 3.0e11 n/(cm² s) | assumption | 3.0×10¹³ n/(cm² s) at 100 %. |
| βeff effective delayed-neutron fraction | 0.0065 – | assumption | Generic thermal LWR; sets the dollar conversion. |
The default flux of 3.0×1013 n·cm−2·s−1 at 100 % and Σf = 0.100 cm−1 imply 3.0e12 fissions·cm−3·s−1, i.e. about 96.1 W·cm−3 at 200 MeV per fission: a coherent representative fuel-region scaling, not a mapping from rated power to a core-average flux. The Xe-135 depletion and worth cross sections are equal by default so that a thermal point value is not silently mixed with a spectrum average.
Table 8.1 — Sensitivity of the default scram case to the evaluated-data variants discussed in §10 (generated from the code; all other values default).
| Case | Peak Xe worth [pcm] | Δ vs. default [%] | at t [h] | Worth at end [pcm] |
|---|
| Default (release 1.0.1 values) | −3976.7 | 0 | 13.342 | −568.1 |
| I-135 half-life 6.57 h (ICRP-107) | −3974.9 | −0.05 | 13.333 | −568.0 |
| I-135 6.55 h, Xe-135 9.168 h (NuDat export) | −3977.0 | 0.01 | 13.325 | −568.2 |
| Yields recomputed from independent yields (Table 10.1, Sn/Sb branching 21 % / 16.6 %) | −3966.1 | −0.27 | 13.342 | −568.5 |
Reasonableness check All model values are unchanged from release 1.0.1, which the prototype values of Table 2.1 do not match; the physically meaningful checks are that the equilibrium xenon worth at rated power is −2119 pcm and the samarium worth −445 pcm, both of the magnitude expected for a thermal LWR, and that the peak xenon worth after the scram scenario is −3977 pcm. These are plausibility checks, not validation.
9. Model Scope and Limitations
Represented The effective I-135 → Xe-135 and Pm-149 → Sm-149 chains with decay and capture; arbitrary piecewise-linear power histories; equilibrium or clean initial states; poison worth in pcm and dollars.
Not represented Axial or radial xenon oscillations and any spatial redistribution; burnup, changing spectrum and cross sections; thermal-hydraulic, Doppler or moderator feedback; control-system response, soluble boron and control-rod worth; direct Sm-149 and Xe-135 explicit precursors beyond the effective yields; flux depression and self-shielding; leakage and adjoint weighting in the worth. The power–flux relation is linear. Results are for engineering insight and training only.
10. Data Provenance and Open Points
What was found The yield chain was re-derived from a community compilation of the ENDF/B-VIII.0 thermal U-235 independent yields [4]: I-135 receives feeding from Sn-135, Sb-135 and Te-135, of which Sn-135 and Sb-135 have delayed-neutron branches (21 % and 16.6 % in the NuDat export [2]). Table 10.1 gives the recomputed cumulative yields with three assumptions for the branching (the Xe bypass column is the independent Xe-135 yield plus 99.7 % of the Xe-135m yield) and compares them with the stored defaults. For mass 149 the independent yields of Pm-149 and Sm-149 are negligible (3.87e−8 and 1.71e−12), so essentially the whole chain yield must be carried by the effective Pm-149 yield, as the model does.
Table 10.1 — Recomputed cumulative yields [atoms/fission] versus stored defaults (consistency check; the stored values are not derived from this calculation).
| Chain / Sn-135, Sb-135 branching | Cum. I-135 | Stored vs. this [%] | Cum. Xe-135 | Xe bypass |
|---|
| none (0 % / 0 %) | 0.0628931 | −0.12 | 0.0654541 | 0.002561 |
| 21 % / 16.6 % (NuDat export [2]) | 0.0626501 | 0.27 | 0.0652111 | 0.002561 |
| 21 % / 22 % | 0.0625714 | 0.40 | 0.0651324 | 0.002561 |
| stored defaults | 0.0628187 | — | 0.065385 | 0.0025663 |
| A = 149: cumulative Pm-149 chain | 0.0108244 | stored 0.0108163: −0.08 % | — | — |
Table 10.2 — Half-lives behind the default decay constants.
| Nuclide | Default λ [s⁻¹] | Default T½ [h] | ICRP-107 [h] [1] | NuDat export [h] [2] |
|---|
| I-135 | 2.9261532e−5 | 6.58 | 6.57 (0.15 %) | 6.55 (0.46 %) |
| Xe-135 | 2.1065742e−5 | 9.14 | 9.14 (0.00 %) | 9.168 (−0.31 %) |
| Pm-149 | 3.6273716e−6 | 53.08 | 53.08 (0 %) | 53.08 (0 %) |
What was not found The primary ENDF/B-VIII.0 files were not read in this release, only the compilation [4] and the NuDat export [2]; the origin of the adopted I-135 half-life of 6.58 h (ICRP-107: 6.57 h, NuDat export: 6.55 h) could not be established. The Pm-149 half-life agrees in all sources and the Xe-135 value agrees with ICRP-107.
Unverified items (1) The stored cumulative yields are reproduced within 0.27 % by the recomputation but not traced to the primary file. (2) The one-group cross sections, Σf, φ1, νf and βeff are assumptions that depend on spectrum, composition and burnup; their effect is not bounded here. (3) The data variants of Table 8.1 change the peak xenon worth by at most 0.27 %, so the unverified half-life and yield details are not the dominant uncertainty.
11. References
External standards, data and books (each entry graded: consulted in full, consulted at abstract/excerpt level, or further reading — not consulted)
- [1] ICRP, Nuclear Decay Data for Dosimetric Calculations, ICRP Publication 107, Ann. ICRP 38 (3), 2008. Values used through the ICRP-107/AME2020/NUBASE2020 dataset of the Python package
radioactivedecay for cross-checking; the publication itself was not read.
- [2] NNDC, NuDat 3, National Nuclear Data Center, Brookhaven National Laboratory (evaluated nuclear structure and decay data, ENSDF). Consulted through a tabular export of half-lives and branching ratios, in full for the nuclides used.
- [3] D. A. Brown et al., “ENDF/B-VIII.0: The 8th Major Release of the Nuclear Reaction Data Library,” Nuclear Data Sheets 148 (2018) 1–142, doi:10.1016/j.nds.2018.02.001. Basis of the thermal U-235 fission yields; primary files not read in this release (§10).
- [4] Community compilation of ENDF/B-VIII.0 independent fission yields and NuDat half-life data (repository “fission-yields-data”, GPL-3.0, T. Ong). Secondary source; the U-235 thermal table was consulted in full for the A = 135 and A = 149 chains. Weaker evidence than the primary library.
- [5] J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, 1976. Poisoning balance equations; further reading — not consulted for this release.
- [6] J. R. Lamarsh and A. J. Baratta, Introduction to Nuclear Engineering, 3rd ed., Prentice Hall, 2001. Xenon and samarium transients; further reading — not consulted for this release.
- [7] E. Hairer, S. P. Nørsett and G. Wanner, Solving Ordinary Differential Equations I, 2nd rev. ed., Springer, 1993. Runge–Kutta and Dormand–Prince methods; further reading.
Antranova model basis
- [A1] Antranova Engineering AB (A. Romas), Xe-135 / Sm-149 Jupyter Notebook prototype, v1.0.0. Historical implementation provenance only (§2).
- [A2] Antranova Engineering AB, Xe-135 / Sm-149 Poisoning Simulator, release 1.0.1. Model basis of the present release (numerics unchanged, §7).