AMPX–NJOY processing map

The AMPX modules used to build a SCALE continuous-energy (CE) library for CANDU lattice physics, and the earlier 56-group covariance library, set beside the NJOY2016 modules that do the same jobs. Each row gives the options used here, the matching NJOY options, and the differences found when the results were compared.

AMPX SCALE 7.0 beta 3 source (scale-public) NJOY 2016.78 Data ENDF/B-VIII.1 Reference for checks official OpenMC ENDF/B-VIII.1 HDF5 library (NJOY-processed) Updated 1 October 2026

The two paths at a glance

AMPX → SCALE CE library (Robus HDF5)

POLIDENT resonance reconstruction, 0 K pointwise data + abbreviated info file
BROADEN Doppler broadening, one temperature at a time merged with PICKEZE + ZEST
PURM + PURM_UP unresolved-resonance probability tables
Y12 secondary distributions: fast, free gas, S(α,β) tabulated, plus 1-D thermal cross sections
TOMATO / ZEST relabel and merge the thermal cross sections
JAMAICAN distributions to sampling PDF/CDF form
PLATINUM one nuclide file, all temperatures + DBRC 0 K elastic
CE_SIMONIZE master library file with links to the nuclide files

NJOY2016 → ACE (→ OpenMC HDF5)

RECONR resonance reconstruction, 0 K PENDF
BROADR Doppler broadening, all temperatures in one step
PURR unresolved-resonance probability tables (UNRESR for Bondarenko factors)
HEATR, GASPR heating and gas production (not needed for k)
THERMR free gas and S(α,β) thermal cross sections and distributions
no separate step: the PENDF tape accumulates each module's output
ACER keeps the ENDF laws (Kalbach-Mann, tabulated) in ACE form
ACER one table per temperature OpenMC: openmc.data merges the tables into one HDF5 file

Step-by-step mapping

Columns: the job, the AMPX module with the options used in this project, the NJOY2016 module and options that do the same job, and what the comparisons showed.

Step	Job	AMPX (options used)	NJOY2016 equivalent	Comparison and differences
1	Read ENDF, reconstruct resonances, linearize	POLIDENT (`polident_new`): `globaleps=1e-3` (mesh tolerance), `globalcombeps=1e-4` (merge with File 3); writes the 0 K TAB1 file and the abbreviated info file (MF1 header, ν̄) that PLATINUM needs.	RECONR: `err=0.001` on card 3; R-matrix-limited (LRF=7) evaluations through the SAMMY-based routines.	R-matrix grid For U-238 (Reich-Moore, 1e-5 eV–20 keV) the 0 K output was byte-identical at `globaleps` 1e-3 and 1e-4, so the tolerance did not refine that grid. O-16 has no resonance parameters (fully pointwise).
2	Doppler broadening	BROADEN: `eps=1e-3`, `t=` one temperature per call, always from the 0 K file; PICKEZE (`temp_sel=1`) keeps only the new temperature's functions; ZEST merges them.	BROADR: `errthn=0.001`, list of temperatures; broadening up to `thnmax`.	sequential error Given several temperatures, BROADEN broadens each from the previous one and the error accumulates: O-16 1/v capture at 0.0253 eV drifted +0.25 % (600 K) and +0.50 % (2500 K) over 15 steps. Directly from 0 K it holds to 0.01 %, matching NJOY. MeV range The NJOY-processed OpenMC library shows the same O-16 2.889 MeV resonance peak at 294, 600 and 1200 K (not broadened above about 1 MeV); AMPX broadens it (2.70 → 2.54 b, 293.6 → 600 K), with integrals equal to 0.007 %.
3	Unresolved-resonance self-shielding (CE)	PURM: Monte Carlo resonance ladders, `nbatch=300 iter=600 nband=20`, all temperatures; PURM_UP adds the File 3 background (`eps=1e-3`); PLATINUM `ptable={file= id=MAT}`.	PURR: ladder sampling, `nbin=20`, `nladr` ladders, all temperatures; ACER carries the tables.	in progress First used for U-238 (unresolved 20–149 keV); comparison with the NJOY tables pending.
3b	Unresolved self-shielding (multigroup)	PRUDE (Bondarenko factors); FABULOUS (full-range Bondarenko tables).	UNRESR (Bondarenko); GROUPR with σ0 values.	not used Multigroup path only.
4	Fast secondary distributions (angle and energy)	Y12 `for=tab`: tabulated double-differential data, `eps=1e-2` (option A), `zap=1 awp=1` (outgoing neutrons only), `id=1` (incident-particle ZA expected by JAMAICAN); reads a copy without MF6 MT18 for VIII.1 actinides.	ACER (iopt=1) translates MF4/MF5/MF6 laws directly into ACE laws; no tabulation tolerance.	different method AMPX re-tabulates every distribution. At Y12 1e-3, H-2 (n,2n) break-up alone is 49 million exit-energy points and conversion takes about 4.5 h, against minutes at 1e-2. Y12 1e-2 against 1e-3: elastic μ̄ within 0.003, mean secondary energy within 0.3 % (median 0.08 %). LAW = −5 Y12 cannot read VIII.1 MF6 MT18 neutron-multiplicity subsections: U-235, U-238 and Pu-239 use MF5/MF4 MT18 instead (recorded modification); Th-232, Pa-231 and Pa-233 have no MF5 MT18 (open).
5	Distribution format for sampling	JAMAICAN (`jamaican_new`): `result=native equi=false eps=1e-3`; converts to lab-frame PDF/CDF (centre-of-mass → lab for elastic).	ACER (laws kept; equiprobable bins only for old formats).	compared Robus rejects equiprobable exit energies (`equi=false`, neutron-only input). JAMAICAN thinning 1e-2 vs 1e-3: μ̄ within 7e-7, Ē′ within 0.05 %.
6	Free-gas thermal scattering	Y12 `free awr= pot= temp=T`, one temperature per call: a 1e-3 pass for the 1-D cross section (`point=`, MT 1007), a 1e-2 pass for the distributions.	THERMR `iin=1` (free gas), `emax`; ACER thermal treatment.	compared O-16 MT 1007 against the exact free-gas formula: median 0.016 %, max 0.14 %, all 15 temperatures (0.10 % at 1e-2). Thermal cutoff 5.05 eV (Y12 default `emax`).
7	Bound thermal scattering S(α,β)	Y12 on MF7, one temperature per call from single-temperature files (`tsl_split.py`, format-only split); 1e-3 cross-section pass + 1e-2 distribution pass.	THERMR `iin=2` (incoherent inelastic), `icoh` (elastic), `emax`; ACER iopt=2 thermal tables. (LEAPR builds MF7 from phonon spectra; not needed.)	283.6–573.6 K D in D₂O MT 1007 against NJOY: median 0.05–0.27 %, within 0.2 % at 0.0253, 0.1 and 1 eV. 600–650 K AMPX 0.5–1.5 % high (median), up to +3.9 % at 1 eV; the excess shrank about 10× from eps 1e-2 to 1e-3 (open). Cutoff 5.05 eV (AMPX) vs 10 eV (OpenMC/NJOY).
8	Manipulate pointwise files	TOMATO (relabel MAT), ZEST (merge, select), PICKEZE (select MAT/MT/temperature), TGEL (sum partials), CHARMIN (format conversion, text dumps for checks), FUNCCALC.	MODER (format); otherwise the PENDF tape is passed from module to module.	no direct equivalent AMPX keeps each step's output as a separate TAB1 file.
9	DBRC data	PICKEZE MT 2 of the 0 K POLIDENT file → PLATINUM `dbrc=` (U-238 and Pu-239).	0 K elastic from RECONR; OpenMC stores it as the "0K" elastic data.	in progress U-238 running.
10	Assemble the nuclide file	PLATINUM (`platinum_new`): `cross=`, `info=`, `kinematics=[…]`, `ptable=`, `dbrc=`, `sigp=`, `zaid=`; all temperatures in one Robus HDF5 nuclide file.	ACER (one ACE table per temperature) → `openmc.data` HDF5.	compared O-16 (15 T) and H-2 in D₂O (17 T) against NJOY: median 0.02–0.03 % at all common temperatures.
11	Library index	CE_SIMONIZE: master HDF5 file with external links (like `ce_v8.0_endf.h5`).	MCNP `xsdir`; OpenMC `cross_sections.xml`.	pending
MG	Multigroup library	JERGENS (weight function), X10 (master library), SIMONIZE, WORM.	GROUPR (`ign`, `iwt`), MATXSR / WIMSR.	used for covariance only
COV	Covariance (SCALE 56 groups)	PUFF-IV: MF33, MF32, MF31, MF35 from a master library; `fix_correlation=false`; COVERX output.	ERRORR (+ COVR / BOXER).	compared Sandwich uncertainties of the two 56-group libraries agree within 3 pcm; 10 (AMPX) vs 19 (NJOY) matrices with \|ρ\| > 1; NJOY ERRORR χ uncertainty is wrong where χ varies inside groups.

Settings used for the CANDU CE library

Setting	AMPX value	Nearest NJOY setting
Cross-section tolerance	1e-3: POLIDENT `globaleps`, BROADEN `eps`, Y12 cross-section passes	RECONR `err`, BROADR `errthn`, THERMR `tol` = 0.001
Distribution tolerance (option A)	Y12 `eps` 1e-2 for distributions; JAMAICAN `eps` 1e-3	none (ACER keeps the ENDF laws)
Temperatures, bound moderators	the kernel's own: D/O in D₂O 283.6–650 K (17); U/O in UO₂ 296–1200 K (8)	THERMR / BROADR at the same temperatures
Temperatures, all other nuclides	293.6, 350, 400, 450, 500, 550, 600, 700, 800, 900, 1000, 1200, 1500, 1800, 2500 K	BROADR temperature list
Thermal cutoff	5.05 eV (Y12 `emax`)	THERMR `emax` (10 eV in the OpenMC library)
Probability tables	PURM 20 bands, 300 × 600 histories	PURR `nbin=20`, `nladr`
DBRC	U-238, Pu-239	0 K elastic in the library (OpenMC: resonance scattering method)
SCALE nuclide IDs	1002 D in D₂O · 2008016 O-16 in D₂O · 1001 H in H₂O · 1008016 O in UO₂ · 1092235/1092238 U in UO₂ · 8001001/8001002 free-gas H/D	OpenMC: thermal tables (c_D_in_D2O, …) attached to nuclides in a material

Differences and open issues

Sequential broadening in BROADEN (fixed in the chain). Several temperatures in one call accumulate an error at low energy that grows with the number of steps: +0.5 % on O-16 capture at 0.0253 eV by 2500 K. Each temperature is now broadened from 0 K.
D₂O kernel at 600–650 K (open). Y12's S(α,β) integration is 0.5–1.5 % high there, though NJOY is smooth in temperature. It is likely linked to Y12's rule that switches to the short-collision-time form whenever a neighbouring S value is zero. A 1e-4 check is not possible on this machine: Y12 needs more than 6 GB per temperature.
Distribution tabulation cost. AMPX re-tabulates every secondary distribution, and its run time grows steeply as Y12's tolerance is tightened. NJOY's ACER has no equivalent cost.
VIII.1 MF6 MT18 (LAW = −5). Y12 cannot read these subsections. Copies without MF6 MT18 are used for Y12 only, and the removal is recorded in modifications.csv. Th-232, Pa-231 and Pa-233 need an MF6 → MF5 conversion before they can be processed.
Thermal cutoff. AMPX uses 5.05 eV and the OpenMC/NJOY library 10 eV. Between 5 and 10 eV, free-atom scattering is used in the AMPX library.
MeV-range broadening. The NJOY-processed reference is not broadened above about 1 MeV (BROADR thnmax), so narrow MeV resonances (O-16 at 2.889 MeV) differ point by point. Integrals agree.
Memory. jamaican_new reads a whole kinematics file and needs about 8× its size in memory. Thermal kernels are therefore processed one temperature at a time.

Sources: the project log ~/ampx-ce/NOTES.md, the comparison outputs ~/ampx-ce/lib/verify_*.txt, and the AMPX source documentation in ~/scale-public/packages/Ampx. NJOY options are from the NJOY2016 manual and the OpenMC processing scripts.