LA-UR-17-20093

The NJOY Nuclear Data Processing System,

Version 2016

Original Author: R. E. MacFarlane

Theoretical Division
Los Alamos National Laboratory

Contributing Authors
D. W. Muir
R. M Boicourt
A. C. Kahler
J. L. Conlin
W. Haeck

Current Editor: A. C. Kahler

Original Issue: December 19, 2016

Updated for NJOY2016.53

November 7, 2019

Abstract
The NJOY Nuclear Data Processing System, version 2016, is a comprehensive com-
puter code package for producing pointwise and multigroup cross sections and related
quantities from evaluated nuclear data in the ENDF-4 through ENDF-6 legacy cardim-
age formats. NJOY works with evaluated files for incident neutrons, photons, and
charged particles, producing libraries for a wide variety of particle transport and reac-
tor analysis codes.
 Disclaimer of Liability:

Neither the United States Government nor the Los Alamos National Security, LLC., nor
any of their employees, makes any warranty, express or implied, including the warranties
of merchantability and fitness for a particular purpose, or assumes any legal liability or
responsibility for the accuracy, completeness, or usefulness of any information, appara-
tus, product, or process disclosed, or represents that its use would not infringe privately
owned rights.

Disclaimer of Endorsement:

Reference herein to any specific commercial products, process, or service by trade name,
trademark, manufacturer, or otherwise, does not necessarily constitute or imply its
endorsement, recommendation, or favoring by the United States Government or the Los
Alamos National Security, LLC. The views and opinions of authors expressed herein do
not necessarily state or reflect those of the United States Government or the Los Alamos
National Security, LLC., and shall not be used for advertising or product endorsement
purposes.

Copyright Notice:

Copyright 2016. Los Alamos National Security, LLC. This software was produced under
U.S. Government contract DE-AC52-06NA25396 for Los Alamos National Laboratory
(LANL), which is operated by Los Alamos National Security, LLC for the U.S. Depart-
ment of Energy. The U.S. Government has rights to use, reproduce, and distribute this
software. NEITHER THE GOVERNMENT NOR LOS ALAMOS NATIONAL SECU-
RITY, LLC MAKES ANY WARRANTY, EXPRESS OR IMPLIED, OR ASSUMES
ANY LIABILITY FOR THE USE OF THIS SOFTWARE. If software is modified to
produce derivative works, such modified software should be clearly marked, so as not
to confuse it with the version available from LANL.
Additionally, redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this list of
conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this
list of conditions and the following disclaimer in the documentation and/or other
materials provided with the distribution.
3. Neither the name of Los Alamos National Security, LLC, Los Alamos National
Laboratory, LANL, the U.S. Government, nor the names of its contributors may
be used to endorse or promote products derived from this software without specific
prior written permission.
THIS SOFTWARE IS PROVIDED BY LOS ALAMOS NATIONAL SECURITY, LLC
AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
IN NO EVENT SHALL LOS ALAMOS NATIONAL SECURITY, LLC OR CONTRIB-
UTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIM-
ITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LI-
ABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.

©Copyright 2016 Los Alamos National Security, LLC All Rights Reserved
LA-UR-17-20093 CONTENTS

Contents

Contents iv

List of Figures x

List of Tables xiii

1 INTRODUCTION 1
1.1 The Modules of NJOY . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Data Flow in NJOY . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Computer Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 History and Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 8

2 NJOY 19
2.1 The NJOY Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Interface Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Free Format Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 ENDF Input-Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Buffered Binary Scratch Storage . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Dynamic Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . 28
2.7 ENDF/B Utility Routines . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8 Math Routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.9 System-Related Utility Routines . . . . . . . . . . . . . . . . . . . . . . 29
2.10 Error and Warning Messages . . . . . . . . . . . . . . . . . . . . . . . . 30
2.11 Coding Details for the NJOY Main Program . . . . . . . . . . . . . . . 33

3 RECONR 43
3.1 ENDF/B Cross Section Representations . . . . . . . . . . . . . . . . . 43
3.2 Unionization and Linearization Strategy . . . . . . . . . . . . . . . . . 45
3.3 Linearization and Reconstruction Methods . . . . . . . . . . . . . . . . 46
3.4 Resonance Representations . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.5 Code Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.6 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.8 Input-Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.9 Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4 BROADR 77
4.1 Doppler-Broadening Theory . . . . . . . . . . . . . . . . . . . . . . . . 77
4.2 Thermal Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3 Data-Paging Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.5 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.6 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
4.7 Input/Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.8 Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

iv NJOY2016
CONTENTS LA-UR-17-20093

5 UNRESR 93
5.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.3 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.4 Output Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.5 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.6 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

6 HEATR 115
6.1 Theory of Nuclear Heating . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.2 Theory of Damage Energy . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.3 Computation of KERMA Factors By Energy Balance . . . . . . . . . . 120
6.3.1 The general case . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.3.2 The special case of fission . . . . . . . . . . . . . . . . . . . . . . 122
6.4 Kinematic Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.5 Computation of Damage Energy . . . . . . . . . . . . . . . . . . . . . . 127
6.6 Heating and Damage from File 6 . . . . . . . . . . . . . . . . . . . . . 130
6.7 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.8 Reading HEATR Output . . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.9 Diagnosing Energy-Balance Problems . . . . . . . . . . . . . . . . . . . 147
6.10 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.11 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.12 Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

7 THERMR 165
7.1 Coherent Elastic Scattering . . . . . . . . . . . . . . . . . . . . . . . . 166
7.2 Incoherent Inelastic Scattering . . . . . . . . . . . . . . . . . . . . . . . 168
7.3 Incoherent Elastic Scattering . . . . . . . . . . . . . . . . . . . . . . . . 174
7.4 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.5 Using the ENDF/B Thermal Data Files . . . . . . . . . . . . . . . . . 179
7.6 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
7.7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
7.8 Input/Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
7.9 Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

8 GROUPR 187
8.1 Multigroup Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8.2 Group Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
8.3 Basic ENDF Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . 190
8.4 Weighting Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.5 Flux Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
8.6 Fission Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
8.7 Diffusion Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . 203
8.8 Cross Sections for Transport Theory . . . . . . . . . . . . . . . . . . . 205
8.9 Photon Production and Coupled Sets . . . . . . . . . . . . . . . . . . . 207
8.10 Thermal Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
8.11 Generalized Group Integrals . . . . . . . . . . . . . . . . . . . . . . . . 211
8.12 Two-Body Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
8.13 Charged-Particle Elastic Scattering . . . . . . . . . . . . . . . . . . . . 217
8.14 Continuum Scattering and Fission . . . . . . . . . . . . . . . . . . . . . 218
8.15 File 6 Energy-Angle Distributions . . . . . . . . . . . . . . . . . . . . . 223

NJOY2016 v
LA-UR-17-20093 CONTENTS

8.16 Smoothing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

8.17 GENDF Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.18 Running GROUPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
8.19 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
8.20 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

9 GAMINR 273
9.1 Description of ENDF/B Photon Interaction Files . . . . . . . . . . . . 273
9.2 Calculational Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
9.3 Integrals Involving Form Factors . . . . . . . . . . . . . . . . . . . . . 276
9.4 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
9.5 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
9.6 I/O Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
9.7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281

10 ERRORR 285
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
10.2 Definitions of Covariance-Related Quantities . . . . . . . . . . . . . . . 287
10.3 Structure of ENDF Files 31, 33, and 40: Energy-Dependent Data . . . 289
10.4 Resonance-Parameter Formats—File 32 . . . . . . . . . . . . . . . . . . 293
10.5 Secondary Particle Angular Distribution Covariances—File 34 . . . . . 297
10.6 Secondary Particle Energy Distribution Covariances—File 35 . . . . . 297
10.7 Radioactive Nuclide Production Covariances–File 40 . . . . . . . . . . 299
10.8 Calculation of Multigroup Fluxes, Cross Sections, and Covariances on
the Union Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
10.9 Basic Strategy for Collapse to the User Grid . . . . . . . . . . . . . . . 306
10.10 Group-Collapse Strategy for Data Derived by Summation . . . . . . . 307
10.11 Processing of Data Derived from Ratio Measurements . . . . . . . . . 310
10.12 Multigroup Processing of Resonance-Parameter Uncertainties . . . . . 314
10.13 Processing of Lumped-Partial Covariances . . . . . . . . . . . . . . . . 315
10.14 Input Instructions and Sample Input for ERRORR . . . . . . . . . . . 315
10.15 ERRORR Output File Specification . . . . . . . . . . . . . . . . . . . . 336
10.16 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
10.17 Input/Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

11 COVR 357
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
11.2 Production of Boxer-Format Libraries . . . . . . . . . . . . . . . . . . . 357
11.3 Generation of Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
11.4 Input Instructions for COVR . . . . . . . . . . . . . . . . . . . . . . . 361
11.5 COVR Example Problem . . . . . . . . . . . . . . . . . . . . . . . . . . 368
11.6 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
11.7 Input/Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
11.8 Retrieval Program for COVR Output Libraries . . . . . . . . . . . . . 372

12 MODER 377
12.1 Code Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
12.2 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
12.3 Sample Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
12.4 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380

vi NJOY2016
CONTENTS LA-UR-17-20093

13 DTFR 383
13.1 Transport Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
13.2 Data Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386
13.3 Plotting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
13.4 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
13.5 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
13.6 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402

14 CCCCR 403
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
14.2 CCCC Procedures and Programming Standards . . . . . . . . . . . . . 404
14.3 The Standard Interface Files . . . . . . . . . . . . . . . . . . . . . . . . 407
14.4 ISOTXS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
14.5 BRKOXS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
14.6 DLAYXS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
14.7 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
14.8 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
14.9 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450

15 MATXSR 453
15.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
15.2 The MATXS Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
15.3 Historical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474
15.4 MATXS Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
15.5 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
15.6 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
15.7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490

16 RESXSR 493
16.1 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
16.2 RESXS Format Specification . . . . . . . . . . . . . . . . . . . . . . . . 494
16.3 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497

17 ACER 499
17.1 ACER and ACE Data Classes . . . . . . . . . . . . . . . . . . . . . . . 499
17.2 Continuous-Energy Neutron Data . . . . . . . . . . . . . . . . . . . . . 500
17.3 Energy Grids and Cross Sections . . . . . . . . . . . . . . . . . . . . . 500
17.4 Two-Body Scattering Distributions . . . . . . . . . . . . . . . . . . . . 504
17.5 Secondary-Energy Distributions . . . . . . . . . . . . . . . . . . . . . . 504
17.6 Energy-Angle Distributions . . . . . . . . . . . . . . . . . . . . . . . . 505
17.7 Photon Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
17.8 Probability Tables for the Unresolved Region . . . . . . . . . . . . . . 511
17.9 Charged-Particle Production . . . . . . . . . . . . . . . . . . . . . . . . 512
17.10 Gas Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
17.11 Consistency Checks and Plotting . . . . . . . . . . . . . . . . . . . . . 513
17.12 Thermal Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
17.13 Dosimetry Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . 519
17.14 Photoatomic Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520
17.15 Photonuclear Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
17.16 Type 1 and Type 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
17.17 Running ACER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522

NJOY2016 vii
LA-UR-17-20093 CONTENTS

17.18 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531

17.19 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545

18 POWR 551
18.1 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551

19 WIMSR 557
19.1 Resonance Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
19.2 Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
19.3 Burn Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
19.4 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
19.5 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569
19.6 WIMS Data File Format . . . . . . . . . . . . . . . . . . . . . . . . . . 572
19.7 WIMSR Auxiliary Codes . . . . . . . . . . . . . . . . . . . . . . . . . . 576
19.8 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576

20 PLOTR 579
20.1 Simple 2-D Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
20.2 Multicurve and Multigroup Plots . . . . . . . . . . . . . . . . . . . . . 583
20.3 Right-Hand Axes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 588
20.4 Plotting Input Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 590
20.5 Three-D Plots of Angular Distributions . . . . . . . . . . . . . . . . . . 591
20.6 Three-D Plots of Energy Distributions . . . . . . . . . . . . . . . . . . 592
20.7 Two-D Spectra Plots from Files 5 and 6 . . . . . . . . . . . . . . . . . 594
20.8 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
20.9 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 603
20.10 Storage Allocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605
20.11 Input and Output Units . . . . . . . . . . . . . . . . . . . . . . . . . . 605
20.12 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606

21 VIEWR 607
21.1 Modular Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
21.2 Using VIEWR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
21.3 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
21.4 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
21.5 The Graphics Module . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
21.6 VIEWR Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

22 MIXR 627
22.1 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
22.2 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
22.3 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632

23 PURR 635
23.1 Sampling from Ladders . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
23.2 Temperature Correlations . . . . . . . . . . . . . . . . . . . . . . . . . 641
23.3 Self-Shielded Heating Values . . . . . . . . . . . . . . . . . . . . . . . . 641
23.4 Random Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
23.5 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
23.6 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
23.7 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650

viii NJOY2016
CONTENTS LA-UR-17-20093

24 LEAPR 653
24.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
24.2 Input Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 669
24.3 LEAPR Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676
24.4 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 716
24.5 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 720

25 GASPR 723
25.1 Gas Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
25.2 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724
25.3 Coding Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725
25.4 Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725

26 NJOY Maintenance and Testing 727

26.1 Code Maintenance with GIT . . . . . . . . . . . . . . . . . . . . . . . . 727
26.2 Standard Test Problems . . . . . . . . . . . . . . . . . . . . . . . . . . 727
26.3 Test Problem 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
26.4 Test Problem 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
26.5 Test Problem 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 736
26.6 Test Problem 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 738
26.7 Test Problem 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
26.8 Test Problem 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 740
26.9 Test Problem 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743
26.10 Test Problem 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 745
26.11 Test Problem 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746
26.12 Test Problem 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749
26.13 Test Problem 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 750
26.14 Test Problem 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 752
26.15 Test Problem 13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754
26.16 Test Problem 14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 756
26.17 Test Problem 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757
26.18 Test Problem 16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 760
26.19 Test Problem 17 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763
26.20 Test Problem 18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 766
26.21 Test Problem 19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769
26.22 Test Problem 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 770
26.23 Application of the NJOY System . . . . . . . . . . . . . . . . . . . . . 772

References 777

Index 787

NJOY2016 ix
LA-UR-17-20093 LIST OF FIGURES

List of Figures
1 RECONR reconstructed xs with smooth, RR and URR regions . . . . . 44
2 Inverted stack mesh generation description . . . . . . . . . . . . . . . . . 47
3 19
F elastic scattering Legendre coefficients from RML data . . . . . . . 63
4 The 10 B (n,α) cross section versus Doppler broadening temperature . . 81
5 The nat C elastic scattering cross section versus Doppler broadening tem-
perature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6 The 240 Pu low energy (n,γ) cross section versus Doppler broadening tem-
perature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7 Energy grid variation with Doppler broadening . . . . . . . . . . . . . . 83
8 Components of nuclear heating . . . . . . . . . . . . . . . . . . . . . . . 116
9 Sample recoil energy and lattice displacement data . . . . . . . . . . . . 120
10 Components of radiation damage energy production for 27 Al . . . . . . 130
11 Total photon energy production and kinematic limits for 55 Mn . . . . . 145
12 MT301 and MT443 for 59 Co . . . . . . . . . . . . . . . . . . . . . . . . . 148
13 Example of File 3 and File 13 energy grid mis-match . . . . . . . . . . . 149
14 Example of energy-balance problems . . . . . . . . . . . . . . . . . . . . 149
15 Computed photon energy production and kinematic values . . . . . . . . 150
16 Example of coherent elastic cross section for a crystalline material (graphite)167
17 Adaptive reconstruction of emission spectra (graphite) . . . . . . . . . . 171
18 Neutron distribution for incoherent inelastic scattering from graphite (T
= 293.6K) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
19 Neutron distribution for incoherent inelastic scattering from graphite (T
= 293.6K), expanded view . . . . . . . . . . . . . . . . . . . . . . . . . . 172
20 Distributions for H in H2 O with E-µ-E ′ ordering . . . . . . . . . . . . . 173
21 Incoherent inelastic distribution for H-H2 O (expanded view) . . . . . . . 183
22 Pointwise and multigroup cross section comparison . . . . . . . . . . . . 192
23 The self-shielding effect on the first three 238 U capture resonances. . . . 196
24 238
U capture resonance integral versus temperature and background cross
section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
25 Flux model predictions for 238 UO2 in water in the region near the 6.7 eV
238
U resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
26 Coupled Neutron/Photon Tables . . . . . . . . . . . . . . . . . . . . . . 208
27 Bragg edges seen in the BeO coherent elastic scattering cross section . . 209
28 Interpolation along lines of constant energy transfer . . . . . . . . . . . 211
29 Sample two-body scattering feed function . . . . . . . . . . . . . . . . . 216
30 Coordinate mapping between CM and LAB reference frames for A=2 . 225
31 GROUPR weight functions on a logarithmic flux/unit lethargy scale . . 236
32 GROUPR weight functions on a linear energy scale . . . . . . . . . . . . 237
33 Coupled neutron-proton-photon table . . . . . . . . . . . . . . . . . . . 251
34 Photon coherent scattering angular distributions . . . . . . . . . . . . . 277
35 Photon interaction cross sections for uranium . . . . . . . . . . . . . . . 282
36 Photon interaction cross sections for hydrogen . . . . . . . . . . . . . . . 282
37 ENDF/B-VII.0 Uranium photoelectric subshell cross sections . . . . . . 283
38 ENDF/B-VII.0 10 B(n,α) covariance data . . . . . . . . . . . . . . . . . . 286
39 Example of elastic scattering and (n,n1 ) covariance data . . . . . . . . . 296
40 Example of angular distribution covariance data . . . . . . . . . . . . . 298
41 252
Cf(n,f) spontaneous fission spectrum covariance data . . . . . . . . . 300
42 Radioactive nuclide production covariance example . . . . . . . . . . . . 301
43 Illustration of energy grid relations. . . . . . . . . . . . . . . . . . . . . . 302

x NJOY2016
LIST OF FIGURES LA-UR-17-20093

44 Ratioed covariance data for 239 Pu(n,f) and 241 Am(n,f) cross sections . . 313
45 7
Li “Lumped" covariance data . . . . . . . . . . . . . . . . . . . . . . . . 316
46 Illustration of Boxer format . . . . . . . . . . . . . . . . . . . . . . . . . 360
47 Pattern-search logic in used in COVR’s subroutine matshd . . . . . . . 362
48 DTFR cross section plot example . . . . . . . . . . . . . . . . . . . . . . 392
49 DTFR cross section plot example, expanded energy range . . . . . . . . 393
50 DTFR scattering matrix plot example . . . . . . . . . . . . . . . . . . . 393
51 DTFR photon production matrix plot example . . . . . . . . . . . . . . 394
52 DTFR plot example, multiple plots per frame . . . . . . . . . . . . . . . 394
53 DTFR plot example, compound edit . . . . . . . . . . . . . . . . . . . . 398
54 Principal cross sections for ENDF/B-VII.0 27 Al . . . . . . . . . . . . . . 515
55 Neutron energy distributions from 27 Al(n,n’α) reaction . . . . . . . . . . 516
56 Neutron scattering distribution from H in H2 O . . . . . . . . . . . . . . 519
57 Sample 2-D plot with default axes . . . . . . . . . . . . . . . . . . . . . 581
58 Sample 2-D plot with modified axes . . . . . . . . . . . . . . . . . . . . 582
59 Sample 2-D plot with linear axes . . . . . . . . . . . . . . . . . . . . . . 583
60 Sample plot with pointwise and multigroup data . . . . . . . . . . . . . 585
61 Sample plot displaying self-shielded cross sections . . . . . . . . . . . . . 586
62 Sample plot with left- and right-hand axes defined . . . . . . . . . . . . 588
63 Sample plot with user data . . . . . . . . . . . . . . . . . . . . . . . . . 591
64 Sample 3-D plot of angular distribution data . . . . . . . . . . . . . . . 593
65 Sample 3-D plot of neutron secondary-energy distribution data . . . . . 594
66 Sample 2-D plot of neutron secondary-energy distribution data . . . . . 596
67 Sample region fill options in VIEWR . . . . . . . . . . . . . . . . . . . . 618
68 Sample 3-D plot with default axis and perspective . . . . . . . . . . . . 619
69 Sample total cross section probability distributions . . . . . . . . . . . . 638
70 Bondarenko-style self-shielding . . . . . . . . . . . . . . . . . . . . . . . 639
71 Self-shielding variation with temperature and dilution . . . . . . . . . . 640
72 Graphite phonon frequency spectrum . . . . . . . . . . . . . . . . . . . . 680
73 Coherent elastic scattering cross section for graphite . . . . . . . . . . . 680
74 Incoherent elastic scattering cross section for graphite . . . . . . . . . . 681
75 Frequency spectra for BeO . . . . . . . . . . . . . . . . . . . . . . . . . . 683
76 The frequency spectrum for H in H2 2O . . . . . . . . . . . . . . . . . . 684
77 S(α, −β) for H in H2 O at room temperature . . . . . . . . . . . . . . . 693
78 S-hat vs beta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694
79 Incoherent inelastic spectra for H in H2 O . . . . . . . . . . . . . . . . . 694
80 Incoherent inelastic spectra for H in H2 O, detailed view . . . . . . . . . 695
81 The incoherent inelastic cross section for H in H2 O at two temperatures 695
82 The incoherent inelastic cross section for H in H2 O for higher incident
energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696
83 The average scattering cosine for H in H2 O compared to the static value
for scattering from atoms at rest . . . . . . . . . . . . . . . . . . . . . . 697
84 A perspective view of an angle-energy distribution for H in H2 O. . . . . 698
85 A perspective view of the isotropic part of the incoherent inelastic scat-
tering from H in H2 O. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 698
86 Comparison of the ENDF/B-VII.0 thermal cross section for water at
lower incident energies with experimental results . . . . . . . . . . . . . 699
87 Comparison of the ENDF/B-VII.0 thermal cross section for water at
higher incident energies with experimental results . . . . . . . . . . . . . 700
88 Harker-Brugger frequency spectrum used for solid methane . . . . . . . 701
89 S(α, −β) for solid methane . . . . . . . . . . . . . . . . . . . . . . . . . 704

NJOY2016 xi
LA-UR-17-20093 LIST OF FIGURES

90 Inelastic and incoherent elastic cross sections for solid methane . . . . . 704
91 Outgoing neutron spectra for solid methane . . . . . . . . . . . . . . . . 705
92 Agrawal-Yip frequency spectrum for liquid methane . . . . . . . . . . . 707
93 Frequency spectrum for liquid methane with translational and rotational
modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707
94 S(α, β) for Liquid Methane . . . . . . . . . . . . . . . . . . . . . . . . . 709
95 Cross section for liquid methane at 100K . . . . . . . . . . . . . . . . . . 710
96 Neutron Spectra for liquid methane . . . . . . . . . . . . . . . . . . . . . 710
97 Keinert-Sax frequency spectrum . . . . . . . . . . . . . . . . . . . . . . 711
98 The static structure factor, S(κ), for liquid hydrogen . . . . . . . . . . . 712
99 script-S vs beta for para-hydrogen . . . . . . . . . . . . . . . . . . . . . 712
100 script-S vs alpha for para-hydrogen . . . . . . . . . . . . . . . . . . . . . 713
101 script-S vs beta for ortho-hydrogen . . . . . . . . . . . . . . . . . . . . . 713
102 script-S vs alpha for ortho-hydrogen . . . . . . . . . . . . . . . . . . . . 714
103 Para-hydrogen neutron spectra . . . . . . . . . . . . . . . . . . . . . . . 714
104 Ortho-hydrogen neutron spectra . . . . . . . . . . . . . . . . . . . . . . 715
105 Liquid hydrogen cross sections . . . . . . . . . . . . . . . . . . . . . . . 715

xii NJOY2016
LIST OF TABLES LA-UR-17-20093

List of Tables
1 Energy Parameter for Effective Doppler-Broadening . . . . . . . . . . . 79
2 Atomic Displacement Energy Data for DPA . . . . . . . . . . . . . . . . 120
3 ENDF/B-III Thermal Data Files . . . . . . . . . . . . . . . . . . . . . . 180
4 ENDF/B-VII Thermal Data Files . . . . . . . . . . . . . . . . . . . . . . 180
5 Covariance Matrices Affected by Ratio Measurements In ENDF/B-V . . 312
6 Reaction pairs in the ENDF/B-V 238 U evaluation . . . . . . . . . . . . . 326
7 Organization of Data for One Group in a Transport Table . . . . . . . . 383
8 Example of a Transport Table with Internal Edits . . . . . . . . . . . . 385
9 Predefined Edits for DTFR . . . . . . . . . . . . . . . . . . . . . . . . . 387
10 Example of DTFR Transport Tables Using Separate Edits . . . . . . . . 388
11 Thermal Reactions Available to DTFR . . . . . . . . . . . . . . . . . . . 391
12 Standard MATXSR particle names . . . . . . . . . . . . . . . . . . . . . 455
13 Standard MATXSR data-type names . . . . . . . . . . . . . . . . . . . . 456
14 MATXSR neutron emitting reaction names . . . . . . . . . . . . . . . . 457
15 MATXSR breadkup reaction (LR flag) names . . . . . . . . . . . . . . . 457
16 MATXSR neutron absorption reaction names . . . . . . . . . . . . . . . 458
17 MATXSR fission reaction names . . . . . . . . . . . . . . . . . . . . . . 458
18 Special MATXSR NJOY names . . . . . . . . . . . . . . . . . . . . . . . 459
19 MATXSR gas production names . . . . . . . . . . . . . . . . . . . . . . 459
20 MATXSR incident proton reaction names . . . . . . . . . . . . . . . . . 460
21 MATXSR thermal material names (ENDF/B-VII) . . . . . . . . . . . . 460
22 MATXSR photoatomic cross section names . . . . . . . . . . . . . . . . 461
23 ACE Data Classes and ZAID suffixes . . . . . . . . . . . . . . . . . . . . 499
24 Example of union grid size variation in ACER .c files . . . . . . . . . . . 501
25 ENDF/B-VII thermal MT numbers used in ACER and THERMR . . . 528
26 ACE particle codes for photonuclear files . . . . . . . . . . . . . . . . . . 544
27 Sample Intermediate Resonance λ Values . . . . . . . . . . . . . . . . . 560
28 Conventional thermal material MT numbers used in WIMSR, GROUPR
and THERMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
29 Oscillator beta values for α = 1 for H in H2 O . . . . . . . . . . . . . . . 693
30 H in H2 O SCT effective temperatures . . . . . . . . . . . . . . . . . . . 697
31 Criticality benchmark results from MCNP5 with NJOY processed
ENDF/B-VII.0 data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 773
32 “FAST" criticality benchmark results from TRANSX/PARTISN with
NJOY processed ENDF/B-VII.0 data . . . . . . . . . . . . . . . . . . . 774
33 “THERMAL" criticality benchmark results from TRANSX/PARTISN
and NJOY processed ENDF/B-VII.0 data . . . . . . . . . . . . . . . . . 774

NJOY2016 xiii
LA-UR-17-20093 LIST OF TABLES

xiv NJOY2016
1 INTRODUCTION LA-UR-17-20093

1 INTRODUCTION
The NJOY nuclear data processing system[1, 2, 3, 4, 5, 6] is a comprehensive
computer code package for producing pointwise and multigroup nuclear cross sec-
tions and related quantities from evaluated nuclear data in the ENDF format.
The U.S. Evaluated Nuclear Data Files (ENDF) have progressed through a num-
ber of versions, notably ENDF/B-III, ENDF/B-IV, ENDF/B-V, ENDF/B-VI,
and ENDF/B-VII[7, 8]. The ENDF format has also evolved through many ver-
sions. Variations of the format called “ENDF-6” were used for ENDF/B-VI and
ENDF/B-VII, and will be used in ENDF/B-VIII. The latest version of the for-
mat is described in the ENDF-102 document[9, 10]. The ENDF format is
also used in other nuclear data libraries such as the JEFF libraries in Europe
and the JENDL libraries in Japan, or in specialized libraries distributed through
the Nuclear Data Section of the International Atomic Energy Agency (IAEA).
These libraries represent the underlying nuclear data from a physics viewpoint,
but practical calculations usually require special libraries for particle transport
codes or reactor core physics codes. This is the mission of NJOY — to take the
basic data from the nuclear data library and convert it into the forms needed for
applications.

1.1 The Modules of NJOY

The NJOY code consists of a set of main modules, each performing a well-defined
processing task. Each of these main modules is essentially a separate computer
program. They are linked to one another by input and output files. The main
modules are supported by a number of subsidiary modules providing things like
physics constants, utility routines, and mathematics subroutines that can be
“used” by the main modules. The NJOY modules are grouped as follows:

NJOY directs the flow of data through the other modules. Subsidiary modules
for locale, ENDF formats, physics constants, utility routines, and math
routines are grouped with the NJOY module for descriptive purposes.
RECONR reconstructs pointwise (energy-dependent) cross sections from ENDF
resonance parameters and interpolation schemes.
BROADR Doppler-broadens and thins pointwise cross sections.
UNRESR computes effective self-shielded pointwise cross sections in the unre-
solved energy range.
HEATR generates pointwise heat production cross sections (neutron KERMA
factors) and radiation damage production cross sections.

NJOY2016 1
LA-UR-17-20093 1 INTRODUCTION

THERMR produces cross sections and energy-to-energy matrices for free or

bound scatterers in the thermal energy range.
GROUPR generates self-shielded multigroup cross sections, group-to-group
scattering matrices, photon production matrices, and charged-particle multi-
group cross sections from pointwise input.
GAMINR calculates multigroup photoatomic cross sections, photon KERMA
factors, and group-to-group photon scattering matrices.
ERRORR computes multigroup covariance matrices from pointwise covariance
data.
COVR reads the output of ERRORR and performs covariance plotting and
output formatting operations.
MODER converts ENDF “tapes” back and forth between formatted (that is,
ASCII) and blocked binary modes.
DTFR formats multigroup data for transport codes that use formats based on
the DTF-IV code.
CCCCR formats multigroup data for the CCCC standard files ISOTXS, BRKOXS,
and DLAYXS.
MATXSR formats multigroup data for the newer MATXS cross-section inter-
face file, which works with the TRANSX code to make libraries for many
particle transport codes.
RESXSR prepares pointwise cross sections in a CCCC-like format for thermal
flux calculators.
ACER prepares libraries in ACE format for the Los Alamos continuous-energy
Monte Carlo MCNP and MCNPX codes. The ACER module is supported
by subsidiary modules for the different classes of the ACE format.
POWR prepares libraries for the EPRI-CELL and EPRI-CPM codes.
WIMSR prepares libraries for the thermal reactor assembly codes WIMS-D
and WIMS-E.
PLOTR makes plots of cross sections and perspective plots of distributions for
both pointwise and multigroup data by generating input for the VIEWR
module.
VIEWR converts plotting files produced by the other modules into high-quality
color Postscript plots.
MIXR is used to combine cross sections into elements or mixtures, mainly for
plotting.
PURR is used to prepare unresolved-region probability tables for the MCNP
continuous-energy Monte Carlo code.
LEAPR produces thermal scattering data in ENDF-6 File 7 format that can
be processed using the THERMR module.
GASPR generates gas-production cross sections in the pointwise PENDF for-
mat from basic ENDF cross sections.

2 NJOY2016
1 INTRODUCTION LA-UR-17-20093

The methods used in these modules and instructions on how to use them are
given in subsequent sections of this report. The sections on the modules are
followed by an additional section on NJOY maintenance and testing.

1.2 Data Flow in NJOY

The modules of NJOY can be linked in a number of different ways to prepare
libraries for various nuclear applications. The following brief summary illustrates
the general flow of data in the code.
RECONR reads an ENDF file and produces a common energy grid for all
reactions (the union grid) such that all cross sections can be obtained to within
a specified tolerance by linear interpolation. Resonance cross sections are recon-
structed using a method of choosing the energy grid that incorporates control
over the number of significant figures generated and a resonance integral criterion
to reduce the number of grid points generated for some materials. Summation
cross sections (for example, total, inelastic) are reconstructed from their parts.
The resulting pointwise cross sections are written onto a “point-ENDF” (PENDF)
file for future use. BROADR reads a PENDF file and Doppler-broadens the data
using the accurate point-kernel method. The union grid allows all resonance re-
actions to be broadened simultaneously, resulting in a saving of processing time.
After broadening and thinning, the summation cross sections are again recon-
structed from their parts. The results are written out on a new PENDF file for
future use. UNRESR produces effective self-shielded pointwise cross sections,
versus energy and background cross section, in the unresolved range. This is
done for each temperature produced by BROADR, using the average resonance
parameters from the ENDF evaluation. The results are added to the PENDF
file using a special format.
If desired, additional special kinds of data can be added to the PENDF file.
HEATR computes energy-balance heating, KERMA, and damage energy using
reaction kinematics or by applying conservation of energy. The ENDF photon
production files can be used in this step, when available. Comparisons of mo-
mentum and energy calculations with the photon files can be used to find energy-
balance errors in the evaluations. For ENDF-6 formatted data, charged-particle
distributions in File 6 are used directly to compute accurate heating and damage
parameters. The energy-balance heating results are added to the PENDF file
using ENDF reaction numbers in the 300 series; kinematic KERMA uses the spe-
cial identifier 443, and damage results use the special identifier 444. THERMR
produces pointwise cross sections in the thermal range. Energy-to-energy inco-

NJOY2016 3
LA-UR-17-20093 1 INTRODUCTION

herent inelastic scattering matrices can be computed for free-gas scattering or

for bound scattering using a precomputed scattering law S(α,β) in ENDF for-
mat. The secondary angle and energy grids are determined adaptively so as to
represent the function to a desired precision by linear interpolation; the angular
representation is then converted to one based on equally probable angles. See
the chapter on THERMR for another possible representation. Coherent-elastic
scattering from crystalline materials can be computed using internal lattice infor-
mation, or for ENDF-6 formatted files, using data from the evaluation directly.
The scheme used provides a detailed representation of the delta functions in
energy and angle. Incoherent-elastic scattering for hydrogenous materials is rep-
resented using equally probable angles computed analytically or using ENDF-6
parameters. The results for all the processes are added to the PENDF file using
special formats and special reaction numbers between 221 and 250. Additional
reactions describing the production of the gases 1 H, 2 H, 3 H, 3 He, and 4 He can
be added to the PENDF file with MT=203 – 207 using the GASPR module.
GROUPR processes the pointwise cross sections produced by the modules
described above into multigroup form. The weighting function for group aver-
aging can be taken to be the Bondarenko form, or it can be computed from the
slowing-down equation for a heavy absorber in a light moderator. Self-shielded
cross sections, scattering matrices, photon production matrices, charged-particle
matrices, and photonuclear matrices are all averaged in a unified way, the only
difference being in the function that describes the “feed” into a secondary group
g ′ with Legendre order ℓ from initial energy E. The feed function for two-
body scattering is computed using a center-of-mass (CM) Gaussian integration
scheme, which provides high accuracy even for small Legendre components of
the scattering matrix. Special features are included for delayed neutrons, cou-
pled energy-angle distributions (either from THERMR or from ENDF-6 eval-
uations using File 6), discrete scattering angles arising from thermal coherent
reactions, and charged-particle elastic scattering. Prompt fission is treated with
a full group-to-group matrix. The results are written in a special “groupwise-
ENDF” format (GENDF) for use by the output formatting modules. GAMINR
uses a specialized version of GROUPR to compute photoatomic cross sections
and group-to-group matrices. Coherent and incoherent atomic form factors are
processed in order to extend the useful range of the results to lower energies.
Photon heat production cross sections are also generated. The results are saved
on a GENDF file.
The covariance module, ERRORR, can either produce its own multigroup

4 NJOY2016
1 INTRODUCTION LA-UR-17-20093

cross sections using the methods of GROUPR or start from a precomputed set.
The cross sections and ENDF covariance data are combined in a way that in-
cludes the effects of deriving one cross section from several others. Special fea-
tures are included to process covariances for data given as resonance parameters
or ratios (for example, fission ν̄). It is also possible to process covariances for the
P1 component of an angular distribution, a secondary-energy distribution, and
radionuclide production. The COVR module uses the output from ERRORR to-
gether with the VIEWR module to make publication-quality plots of covariance
data; it also provides output in the efficient BOXER format, and it provides a site
for user-supplied routines to prepare covariance libraries for various sensitivity
systems.
MODER is often used at the beginning of an NJOY job to convert ENDF
library files into binary mode for calculational efficiency, or at the end of a job to
obtain a printable version of a result from ENDF, PENDF, GENDF or ERRORR
input. It can also be used to extract desired materials from a multimaterial
library, or to combine several materials into new ENDF, PENDF, GENDF or
ERRORR files. DTFR is a simple reformatting code that produces cross-section
tables acceptable to many discrete-ordinates transport codes. It also converts
the GROUPR fission matrix to χ and ν̄σf and prepares a photon-production
matrix, if desired. The user can define edit cross sections that are any linear
combination of the cross sections on the GENDF file. This makes complex edits
such as gas production possible. DTFR also produces plotting files for VIEWR
to use in making routine plots for the cross sections, P0 scattering matrix, and
photon production matrix. This module has become somewhat obsolete with
the advent of the MATXS/TRANSX system.
A number of other interface file formats are available from NJOY. The CC-
CCR module is a straightforward reformatting code that supports all the options
of the CCCC-IV[11] file specification. In the cross-section file (ISOTXS), the
user can choose either isotope χ matrices or isotope χ vectors collapsed using
any specified flux. The BRKOXS file includes the normal self-shielding factors
plus self-shielding factors for elastic removal. The DLAYXS provides delayed-
neutron data for reactor kinetics codes. Note that some of the cross sections
producible with NJOY are not defined in the CCCC files. For that reason, we
have introduced the new CCCC-type material cross section file MATXS. The
MATXSR module reformats GENDF data for neutrons, photons, and charged
particles into the MATXS format, which is suitable for input to the TRANSX
(transport cross section) code[12]. TRANSX can produce libraries for a variety of

NJOY2016 5
LA-UR-17-20093 1 INTRODUCTION

particle transport codes that have been used over the years, such as ANISN[13],
ONEDANT[14], TWODANT[15], and DIF3D[16]. TRANSX has special features
to support the latest SN transport code from Los Alamos, PARTISN[17]. The
MATXS format uses efficient packing techniques and flexible naming conventions
that allow it to store all NJOY data types. A companion module, RESXSR, for-
mats pointwise data into a CCCC-like format for use in thermal flux calculators.
Pointwise data can also be fed directly into the ACER module. This module
prepares cross sections and scattering laws in ACE format (A Compact ENDF)
for the MCNP code[18]. All the cross sections are represented on a union grid for
linear interpolation by taking advantage of the representation used in RECONR
and BROADR. “Laws” for describing scattering and photon production are very
detailed, providing a faithful representation of the ENDF-format evaluation with
few approximations. The data are organized for random access for purposes of
efficiency. MCNP handles self-shielding in the unresolved-energy range using
probability tables. The PURR module of NJOY can be used to prepare these
tables and add them to a PENDF file for reading by the ACER module.
Another alternate path for multigroup data is to use the POWR module
to produce libraries for the power reactor codes EPRI-CELL or EPRI-CPM 1 .
Similarly, the WIMSR module can be used to prepare libraries for the thermal
reactor assembly codes WIMS-D and WIMS-E[19].
At the end of any NJOY run, the PLOTR and VIEWR modules can be used
to view the results or the original ENDF data. PLOTR can prepare 2-D plots
with the normal combinations of linear and log axes, including legend blocks
or curve tags, titles, and so on. Several curves can be compared on one plot
(for example, pointwise data can be compared with multigroup results), and
experimental data points with error bars can be superimposed, if desired. Plots
can be prepared showing the ratio or percent difference of two cross sections.
PLOTR can also produce 3-D perspective plots of ENDF and GENDF angle
or energy distributions or thermal S(α, β) tables. The output of PLOTR is
passed to VIEWR, which renders the plotting commands into high-quality color
Postscript graphics for printing or for viewing. The HEATR, COVR, DTFR,
and ACER modules also produce plotting files in VIEWR format that are useful
for quality reviews of data-processing results. The MIXR module can be used to
combine isotopes into elements for plotting and other purposes. It only works
for simple cross sections at the present time.
1
EPRI-CELL and EPRI-CPM are proprietary products of the Electric Power Research Institute (EPRI),
3420 Hillview Avenue, Palo Alto, CA 94304.

6 NJOY2016
1 INTRODUCTION LA-UR-17-20093

1.3 Computer Implementation

NJOY2016 is written in a modern subset of Fortran-90 and later, stressing the use
of modules. The modules help to enforce information hiding. Thus, for example,
the RECONR module only makes the single subroutine reconr public; all the
internal routines and data structures of the module are protected. The modules
also help to promote logical structuring. For example, all of the routines and data
structures for working with the ENDF formats, such as subroutine contio and
the math,mfh,mth values are made public by the endf module, and they can be
easily accessed wherever they are need by the use endf statement. As another
example, what could be a very large ACER module is made more structured by
providing subsidiary modules for handling each of the different “classes” of the
ACE format (e.g. acefc for continuous class “c” data, or acepn for “u” class
photonuclear data).
The advanced capabilities that modern Fortran provides for typing vari-
ables (the “kind” property) have enabled us to remove all of the complex short-
word/long-word controls from NJOY. Almost all of the internal data in NJOY
are handled using 8-byte kinds for real and integer words. These properties are
set up in the locale module, and they can be changed, if necessary, without
touching the balance of the code. An exception is made for the CCCC modules
CCCCR, MATXSR, and RESXSR, where 4-byte variables and equivalencing are
still used to construct records with mixed real, integer, and Hollerith values.
Modern Fortran also provides a capability for the dynamic allocation of mem-
ory for data structures. In the past, NJOY used its own storag system for
this purpose; this system has now been abandoned for the readability, consis-
tency, and exportability provided by the new Fortran standard. We have limited
ourselves to the use of “allocatable arrays”. This relieves us of some storage
limitations in NJOY, and introduces others.
The earliest versions of NJOY used their own free-form input routine, FREE,
which was developed long before Fortran supported equivalent capabilities. More
recently we have abandoned FREE in favor of the standard Fortran READ*
method. In doing this, we have lost some capabilities (such as the repeat field),
but we have gained in transportability. Test fields that were previously delimited
with the star character now must be delimited with the single-quote character.
For consistency and convenience, the NJOY modules in previous versions of
the code made use of a set of common functions and subroutines located in
the NJOY module. Beginning with NJOY2012, these utility routines have been
repackaged into Fortran-90 modules. They include locale for localization vari-

NJOY2016 7
LA-UR-17-20093 1 INTRODUCTION

ables, physics for physics constants, mainio for input, output, and scratch units,
util for utility routines like time and date, endf for ENDF processing routines
and variables, and math for mathematics routines. Each of these modules only
makes public the minimum set of routines and variables needed by the other
modules that “use” them. These modules are described in detail in the NJOY
chapter of this report.
NJOY is heavily commented. Each module starts with a long block of com-
ment cards that gives a brief description of the module and then gives the current
user input instructions. Users should always check the input instructions in the
current version of the NJOY source code rather than the instructions summa-
rized in this manual — changes may have been made. For the convenience of
users, the input instructions are also available on the web[20]. Furthermore, each
function or subroutine starts with a block of comment cards that describes its
function and special requirements. Additional lines of comment cards are used
inside each procedure to block off its major components.
Typography conventions for Fortran differ from place to place. On most
machines at Los Alamos National Laboratory (LANL), Fortran text is given in
lowercase. In order to avoid translation problems, previous versions of NJOY
avoided using mixed-case text for comments or for labels on graphs. We are
gradually moving away from this limiting convention. In this report, Fortran
text and variable names are printed using a lowercase typewriter font. ENDF-
related formats and variables are given using an uppercase TYPEWRITER FONT.
CCCC formats were traditionally given using uppercase characters, but the usage
in this report is mixed.

1.4 History and Acknowledgments

NJOY was started as a successor to MINX[21] (A Multigroup Interpretation
of Nuclear X-sections) late in 1973 (it was called MINX-II then). The current
name was chosen in late 1974 to be evocative of “MINX plus” and to eliminate
the reference to “multigroup.” The first goals were to add a photon production
capability like that in LAPHAN0[22], to add a photon interaction capability like
GAMLEG[23], to provide an easy link to the Los Alamos 30-group libraries of
the day using DTF[24] format, and to merge in the capabilities of ETOPL[25]
to produce libraries for the MCN Monte Carlo code (the ancestor of MCNP).
Most of the work was done by MacFarlane; Rosemary Boicourt joined the project
in 1975. First, the RESEND[26] and SIGMA1[27] modules of MINX were con-
verted to use union grids, and a new method of resonance reconstruction was

8 NJOY2016
1 INTRODUCTION LA-UR-17-20093

developed. These steps led to RECONR and BROADR. UNRESR, which was
based on methods from ETOX[28], was moved over from MINX with only a
few changes. Next, a completely new multigroup averaging program, GROUPR,
was developed around the unifying concept of the “feed function,” which han-
dled neutron- and photon-production cross sections in a parallel manner. The
CM Gaussian integration for discrete two-body scattering was developed. DTFR
was developed as the first NJOY output module. The first versions of the NJOY
utility codes were introduced; the new concepts of “structured programming”
inspired some of the features of the new NJOY code.
Major influences during this period included Don Harris, Raphe LaBauve,
Bob Seamon, and Pat Soran at Los Alamos, and Chuck Weisbin at the Oak
Ridge National Laboratory (ORNL). Odelli Ozer at the Brookhaven National
Laboratory (and later EPRI) helped with RESEND, and Red Cullen at Lawrence
Livermore National Laboratory and John Hancock at Los Alamos helped with
the Doppler broadening module. In those days, the development of NJOY was
supported by the U.S. Fast Breeder Reactor and Weapons Programs.
Code development continued during 1975. The ERRORR module was added
for calculating covariances from ENDF/B files. The ACER module was created
by borrowing heavily from ETOPL and Chuck Forrest’s MCPOINT code. Rich
Barrett joined the project, and he did most of the work in creating a new CCCCR
module for NJOY that had several advances over the MINX version and met the
CCCC-III standards[29]. By the end of the year, HEATR had also been added
to the code (with ideas from Doug Muir). HEATR gave NJOY most of the
capabilities of the original KERMA factor code, MACK [30].
During 1976, free-form input and dynamic data storage were added to NJOY.
GAMINR was written to complete the original NJOY goal of processing photon
interaction cross sections, and the MATXSR module was designed and written,
primarily by Rich Barrett. This completed the capability to construct fully cou-
pled cross sections for neutron-photon heating problems. A major new effort was
writing the THERMR module to improve upon the thermal moderator scatter-
ing cross sections then produced using the FLANGE-II[31] and HEXSCAT[32]
codes, and starting the POWR module to produce cross sections for the EPRI-
CELL and EPRI-CPM codes used by the U.S. electric utility companies. This
work was funded by the Electric Power Research Institute (EPRI).
The first release of NJOY to what was then called the Radiation Shielding
Information Center (RSIC) at Oak Ridge and to the National Energy Software
Center (NESC) at Argonne was NJOY77 in the summer of 1977. This version

NJOY2016 9
LA-UR-17-20093 1 INTRODUCTION

was tested and converted for IBM machines by R. Q. Wright (ORNL). Also,
TRANSX was developed during 1977, the MATXS1 30x12 library was produced
based on ENDF/B-IV, the flux calculator was added to NJOY to support the
EPRI library work, and the first version of the EPRI-CELL library was generated
and used.
A second release of NJOY called NJOY78 was made in 1978[33]. In addi-
tion, further improvements were made for preparing EPRI cross sections, the
MATXS/TRANSX system was improved, and a thermal capability was added
to the MCNP Monte Carlo code using cross sections from THERMR as processed
by ACER.
In 1979, the radiation damage calculation was added to HEATR, and the
GROUPR flux calculator was further improved. In 1980, a plotting option
was added to ERRORR. During this period, NJOY had become more stable.
Changes usually consisted of small improvements or bug fixes instead of major
new capabilities. Starting in this period, NJOY received some support from the
U.S. Magnetic Fusion Energy Program, mostly for covariance work and TRANSX
related library support.
In 1981 and 1982, improvements included the momentum-conservation method
for radiative capture in HEATR . Analytic ψχ broadening was added to RE-
CONR for some cases, and the integral criteria for resonance reconstruction
with significant-figure control were installed in RECONR. Several new capabil-
ities were added to ERRORR, and the COVR module was added to NJOY to
handle both ERRORR plotting and covariance library output. Much of this
covariance-related work was done by Doug Muir. Wiley Davidson (LANL) and
B.H. Broadhead and R. W. Peele at ORNL were helpful. Some support for the
documentation work came from the Paul Scherer Institute (PSI) in Switzerland
and OECD Nuclear Energy agency during visits to those institutions. In addi-
tion, CCCCR was updated to the CCCC-IV standards. A new release, NJOY
(10/81), was made to the code centers, and the first two volumes of a new NJOY
report were written and published. European users began to make important
contributions about this time. Enrico Sartori of the NEA Data Bank, then at
Saclay in France, Margarete Mattes of the University of Stuttgart in Germany,
and Sandro Pelloni of the Paul Scherer Institute in Switzerland deserve mention.
Another major release, NJOY 6/83, was made in 1983. By this time, NJOY
was in use in at least 20 laboratories in the United States and around the world.
Small improvements continued, such as the kinematic KERMA calculation in
HEATR. The temperature dependence of the BROADR energy grid was intro-

10 NJOY2016
1 INTRODUCTION LA-UR-17-20093

duced early in 1984 based on an observation by Ganesan (India). Volume IV of

the NJOY report was published in 1985, and Volume III appeared in 1987.
The next big set of improvements in NJOY was associated with the introduc-
tion of the ENDF-6 format. This required significant changes in RECONR to
support new resonance formats like Reich-Moore and Hybrid R-Function (im-
plemented with help from Charlie Dunford of Brookhaven), in HEATR to im-
plement direct calculations of KERMA and damage from charged-particle and
recoil distributions in File 6, in THERMR to support new formats for File 7,
and in GROUPR to support the group-to-group transfer matrices using energy-
angle data from File 6. The PLOTR module was also developed during this
period. The result was the release of NJOY89[34] in time for processing the new
ENDF/B-VI library and the JEF-2 library (which was also in ENDF-6 format).
During 1989 and 1990, initial processing of ENDF/B-VI and JEF-2 exposed a
number of small problems that had to be fixed. In addition, the ACER module
was rewritten to clean it up, to add capabilities to produce ACE dosimetry
and photoatomic libraries, and to provide for convenient generation of files in
several different formats for users away from LANL. A MIXR module was added
to NJOY, mostly to allow elemental cross sections to be reconstructed from
ENDF/B-VI isotopes for plotting purposes. A new technique was introduced
into GROUPR and all the output modules for multigroup data that provided
for more efficient processing of fission and photon production matrices with lots of
low-energy groups. Major revisions were made to the MATXS format to allow for
charged-particle cross sections, to pack matrices with lots of low-energy groups
more efficiently, and to make inserting and extracting new materials easier. The
WIMSR module, which had been under development for a number of years in
cooperation with WIMS users in Canada and Mexico, was introduced into NJOY.
Finally, the PURR module for generating unresolved-region probability tables
for use with MCNP, which had also been under development for many years,
was formally added to the code. The result of this year-and-a-half of work was
NJOY91.
During the balance of 1991 and 1992, a number of changes were made in
response to problems identified by users as ENDF-6 evaluations began to be
used in earnest. The first of a number of attempts to handle laboratory-frame
Legendre data in File 6 in MCNP was made — namely, the attempt to convert
such sections to use Kalbach systematics. Treatments for phase-space and angle-
energy versions of File 6 were added in 1993, as well as improvements in WIMSR
(with contributions of Fortunato Aguilar, ININ, Mexico). Plotting was added

NJOY2016 11
LA-UR-17-20093 1 INTRODUCTION

for S(α, β) curves from ENDF-6 File 7 and for 2-D and 3-D plots from GENDF
data. HEATR was modified to include incident-energy effects on the fission Q
values in 1993. In 1994, we made another attempt to handle the cases of File 6
using laboratory-frame Legendre expansions for MCNP by converting them into
the LAW=7 angle-energy format. A fix to the Kalbach option for energy-angle
distributions was made based on work by Bob Seamon (LANL). Also, work was
done in HEATR to properly handle the damage energy cutoff at low energies.
Quite a bit of work was done during this period to improve the portability
of the code by installing it on a variety of systems and using codes like cflint.
Other people providing suggestions during this period included Margarete Mattes
(Stuttgart), Piet De Leege (Delft), John White (RSIC), and the Petten users.
The final version of the NJOY91 series was 91.118 in November 1994, and it was
able to process all of the ENDF/B-VI evaluations that had been tried up to that
date. A new user manual for NJOY91[3] was released in 1994. This document
was the primary reference for NJOY until the release of NJOY2012 manual[5].
NJOY94 was issued at the end of 1994 to clean up 91.118 after 3 years of
changes. It also provided a new direct-to-Postscript plotting system by splitting
the old PLOTR module into PLOTR and VIEWR, an updated version of the
PURR module, and a new LEAPR module for computing thermal scattering
functions. In 1995, a new capability was added to handle nuclide production
based on a proposed extension to the ENDF-6 File 8, and the new GASPR mod-
ule was added to handle gas production. Early in 1996, a capability to pass
damage cross sections into MCNP was added. A number of changes were made
based on suggestions from users, and based on processing experience with the
new 150-MeV evaluations becoming available during this period. A change in the
erfc function was made to improve the consistency of 1/v cross sections (prob-
lem observed by Cecil Lubitz, KAPL). Late in the year, the plotting system was
upgraded to promote color Postscript plotting. In mid 1997, a large update to
the WIMSR module was made based on the work of Andrej Trkov (Slovenia).
Problems with calculational accuracy (especially for BROADR) led to a major
upgrade of the math routines. Routines from the SLATEC library were adopted
and converted to NJOY style. A new capability was added to BROADR to com-
pute and display some standard thermal quantities, such as thermal cross sec-
tions, g factors, and ratio integrals (α, K1). After some additional bug patches
and portability improvements, the NJOY94 series was concluded with the is-
suance of 94.105 in July of 1997.
NJOY 97.0 was released in October of 1997. The major change was to move

12 NJOY2016
1 INTRODUCTION LA-UR-17-20093

to using 8-byte words throughout the code. Modern resonance evaluations push
the limit of 6 significant figures available using 32-bit words, and obtaining con-
sistency between people using 32-bit machines and people on 64-bit machines
was becoming more and more desirable. Many changes were associated with
this, including taking great care with all literal constants in the code, removing
Hollerith constants, changing from the FREE input routine to a more standard
READ* method, and developing the techniques to use either 7-digits or 9-digits
in resonance grids, as needed. SAVE statements were added to support stack-
based compilers. The code now automatically determined the version of ENDF
data on the input files, which changed the input a little. Finally, some steps
were made to remove statement numbers and move part way to Fortran-90 style
without abandoning Fortran-77 compatibility.
A number of changes were made during 1998 to cement the features of the
new version. The first new capability added was new conditional probabilities
in PURR for heating. This feature requires running HEATR with partial cross
sections for elastic, fission, and capture. Corresponding changes were also made
in ACER. The tolerances in RECONR and BROADR were modified to use a
tighter tolerance in the thermal range to help preserve the 0.0253-eV cross section
and other thermal parameters better (Lubitz). Piet DeLeege (Delft), Sandro
Pelloni (PSI), and Andrej Trkov (IJS/Slovenia) were helpful in finding problems
during this period. A large number of changes were made based on issues raised
by testing the code with the ftnchek program. These help for portability between
different computer systems. This work was motivated after Giancarlo Panini
(Italy) questioned small differences seen using different compilers. Early in 1999,
the method for setting the damage threshold in HEATR was revised. An internal
table of default values was provided, and a new option to allows the user to enter
a value for Ed was added. This change affects the damage near the threshold for
elastic displacements. John White (RSIC) helped with this. Based on some code
comparisons, Nancy Larson (ORNL) promoted updating the basic constants in
the various codes to enhance compatibility, and we followed suit in NJOY. More
work was done to enhance activation processing and to provide an automatic
loop over all the production reactions for GROUPR. Finally, a new option to
properly handle channel spin in Reich-Moore resonance evaluations was added
using code contributed by ORNL. The NJOY97 series was frozen in September
of 2000.
During the last part of the life of NJOY97, we also had NJOY99 available
(dated 31 December 1999). It is a cleaned up version of NJOY97 that moves

NJOY2016 13
LA-UR-17-20093 1 INTRODUCTION

further toward using block constructs and eliminating statement numbers (but
lots of statement numbers are still left). It is compatible with both Fortran-77
and Fortran-90 compilers. Physical constants were moved into a few common
blocks and standardized on the 1987 CODATA values from NIST. The NJOY
Y2K problem was fixed. The bulk of the changes are in the ACER module
to support new MCNP features, including high-energy data, incident charged
particles, and photonuclear data. Before release, this version was tested for a
large number of different machine/compiler combinations, including tests on X86
PCs. Compatibility has improved to the point where the unix diff function can be
used to compare files and only a small number of very small differences are found.
This version was able to process all the materials from ENDF/B-VI Release 5 into
a library using the new ACE formats (cumulative elastic distributions, LAW=61,
and charged-particle production sections) that were planned to come out with
MCNP4c.
We began making patches to NJOY99 in the spring of 2000, starting with
fixing a problem with the cold hydrogen and deuterium calculations in LEAPR.
This was followed by a number of other small patches. A capability was added
for processing anisotropic charged particle emission in ACER. In 2001 the series
of MT numbers from 875 – 891 was installed to represent levels in the (n,2n)
reaction (needed for a new European 9 Be evaluation). We added a capability
to include delayed neutron data in the ACE files to feed a new capability in
MCNP. A photonuclear capability was added to MATXSR — this enables the
TRANSX code to generate fully coupled sets for n-γ transport. Some coding
was added to generate fluorescence data for MCNP using the existing format
with new numbers coming from the ENDF/B-VI atomic data. This work does
not completely support all the atomic data now available in ENDF/B-VI. The
ACER consistency checks were upgraded to include delayed neutrons, and plots
for ν̄ and the delayed neutron spectra were added. In addition, delayed neu-
tron processing was generalized to allow for 8 time bins as used by the JEFF
evaluations.
PLOTR was modified to allow ratio and difference plots using the right-hand
scale. The default tolerances used in RECONR and BROADR were tightened
up a bit. Some changes to the heating for photoatomic data were provided by
Morgan White (LANL). A change was made to HEATR to provide the photon
contribution to heating using a special MT number (442). When passed to
MCNP, it allows the code to get good answers for heating even when photons are
not being transported. Late in 2002, LEAPR was updated to include coherent

14 NJOY2016
1 INTRODUCTION LA-UR-17-20093

elastic scattering for FCC and BCC crystalline lattices. In 2004, some extensions
to the energy grid used for incoherent inelastic scattering were made. A few
additional smaller patches were also made during this period.
In 2005, changes proposed for THERMR and LEAPR by Margarete Mattes
(IKE/ Stuttgart) to support the new IAEA-sponsored evaluations for thermal
scattering in water, heavy water, and ZrH were installed. Some additional group
structures used in Europe were added. Through this period, we were always
increasing the storage space allowed as we adapted to newer and larger evalua-
tions coming out for ENDF/B-VII. During the summer, a number of changes to
covariance processing were provided by Andrej Trkov (working at IAEA). Early
in 2006, a new sampling scheme for thermal scattering was developed for MCNP
that uses continuous distributions for secondary energy instead of the previous
discrete values. This removed unsightly artifacts in computed fluxes at low en-
ergies, and it alleviated some problems that the cold-neutron-source people were
having. More code improvements were made based on detailed compiler check-
ing. Some errors in the treatment of energy-dependent fission Q were fixed based
on a review of the work of Dave Madland (LANL). An approximate treatment
for the relativistic gamma in the ENDF/B-VII evaluation for n+1 H was added
in HEATR and GROUPR based on theoretical work from Gerry Hale (LANL).
We added a new plot to the ACER set that shows the recoil part of the heating.
This is a sensitive test of energy-balance.
In the summer of 2007, we added some smoothing options to make the
low-energy shape of neutron distributions look more like the theoretical shape,
namely sqrt(E ′ ). Similar changes were provided for delayed neutron spectra.
Additional smoothing was provided for some of the fission spectra at energies
above 10 MeV using an exponential shape. A change was made in GROUPR to
override Cartesian interpolation in favor of unit-base interpolation for scattering
distributions. This gives smoother scattering source functions and is consistent
with what MCNP does.
In 2007, a big change was made to covariance processing by replacing the
original NJOY ERRORR module with ERRORJ[35] as contributed to the NJOY
project by Japan (Go Chiba). This new module added covariance capabilities
for the more modern resolved-resonance representations, angular distributions,
and secondary-energy distributions. A series of additional changes to the new
ERRORR were made over the next couple of years. Go Chiba, Andrej Trkov
(IAEA and IJS Slovenia), and Ramon Arcilla (BNL) were involved in this. A
capability to handle energy-dependent scattering radius data in the unresolved

NJOY2016 15
LA-UR-17-20093 1 INTRODUCTION

range was added. In 2008, ERRORJ work continued. Some work was done in
PLOTR to implement graphs of GROUPR emission spectra. In 2009, changes
were made for unresolved resonance cross sections to force log-log interpolation
to better represent 1/v cross sections. The parameter that looked for steps in the
unresolved-range energy grids that were too large was changed from its former
value of 3 to 1.26. Some of the steps in ENDF/B-VII are unreasonably large when
representing 1/v cross sections. The default energy grid used in the unresolved
range was upgraded to one using about 13 points per decade. Additional pages
were added to the ACER plots to display the unresolved-range self-shielded cross
sections. This work was influenced by Red Cullen (LLNL). In a related effort,
some changes were made in the binning logic for PURR to improve unresolved-
range results. In 2010, changes were made to support the processing of the
IRDF international dosimetry file, including the addition of many new reaction
MT numbers. Some work was done on the photonuclear options to support the
TENDL-2009 library. More ERRORR changes from Trkov were implemented,
and a capability to process the uncertainty in the scattering radius was added.
Finally, several updates were submitted by the Japan Atomic Energy Agency
(JAEA) related to processing discrete photon data from file 6 into ACE files.
NJOY 99.364 was released in February of 2011.
Concurrent with the maintenance of NJOY99, we worked on a new Fortran-
90 version of NJOY that took advantage of the module system, the numerical
precision system, and the dynamic storage system of modern Fortran. This
enabled us to get rid of COMMON statements (always problematic in Fortran)
and to obtain a more readable scheme for allocated storage. Additional work was
done to upgrade to block structures, but some statement numbers still remain.
That new version was called NJOY2010, but was only released to a few in-house
(LANL) users, selected external users at Brookhaven, Oak Ridge and Argonne
National Laboratories plus the Naval Reactors laboratories, Bettis and KAPL
in the United States and to the Atomic Weapons Establishment in the UK.
Further code revisions lead to NJOY2012[5]. For most applications NJOY2012
is compatible with NJOY99. However, there were some new capabilities in
NJOY2012 that are not supported by NJOY99. We borrowed coding from the
SAMMY[36] code (thanks are due to Nancy Larson, ORNL) to handle the new
Reich-Moore-Limited resonance parameter representation now included in the
ENDF-6 format. This coding is used in RECONR to generate pointwise reso-
nance cross sections and angular distributions. It is used in ERRORR to generate
the sensitivity of resonance cross sections to the resonance parameters for use

16 NJOY2016
1 INTRODUCTION LA-UR-17-20093

in calculating cross section uncertainties and covariances in the resonance range.

In THERMR, we added an option to construct thermal cross sections using
(E, µ, E ′ ) ordering in addition to the normal (E, E ′ , µ) ordering. This ordering
is convenient for comparing to experiment, and it is used for thermal sampling
in some Monte Carlo codes.
With NJOY2016 we now have an “open source” code release. The current
version of the source code is now freely available, and can be downloaded from
http://njoy.lanl.gov.

NJOY2016 17
LA-UR-17-20093 1 INTRODUCTION

18 NJOY2016
2 NJOY LA-UR-17-20093

2 NJOY
The NJOY module of the NJOY system contains the main program for the sys-
tem, which directs the sequence of other modules that makes up the desired
calculational path. This section of the report also describes a number of sub-
sidiary modules containing common data structures, subroutines, and functions
for use in the other modules. This manual describes NJOY version 2016 (or
NJOY2016 for short).

2.1 The NJOY Program

The njoy program starts by initializing the page size for blocked-binary ENDF
files, opening the output listing file, and writing the NJOY banner on the output
listing. It then sets up a loop that simply reads a module name and calls the
requested module. The loop continues until the “stop” name is read, and then
NJOY exits. The first card read by any module contains the unit numbers for
the various input and output files. In this way, the output of one module can
be assigned to be the input of another module, thereby linking the modules to
perform the desired processing task. An example of the linking procedure is
given below:

[mount an ENDF file as tape20]

-- Optional comment card (signified by "dash" "dash" "space")
reconr
20 21
[input lines for RECONR]
groupr
20 21 0 22
[input lines for GROUPR]
dtfr
22 23 21
[input lines for DTFR]
stop

Optional comment cards may only appear at the point in an input deck where
a module name is expected. Multiple comment cards are allowed and each must
start with the three character string “dash" “dash" “space". An important feature
of a good modular system is that there be a minimum of interactions between

NJOY2016 19
LA-UR-17-20093 2 NJOY

the modules in order to reduce side effects. In the NJOY system, modules
communicate only by means of the input and output units specified as shown
above, and through a limited number of common constants provided by the
subsidiary modules. The common constants used in the NJOY main program
are in the following modules:

version provides the version number string vers, initially “2016.0,” and the
corresponding date vday, initially “ddmmm16,”
locale provides localization information, including the laboratory and machine
strings, initially “our-lab” and “our-mx,”
mainio provides the system unit numbers, nsysi, nsyso, and nsyse,
endf provides the page size for blocked-binary ENDF files, npage. We recom-
mend users not change this value as there may still be some legacy fixed
array definitions sprinkled in the code that were sized based upon the cur-
rent value.

The input instructions for the NJOY module are given as comment cards at the
beginning of the module. They are reproduced here for the convenience of the
user.

!---input specifications (free format)----------------------------------

!
! card 1 module option
!
! module six character module name, e.g., reconr.
! it is not necessary to use quotes.
!
! repeat card 1 for each module desired, and
! use the name "stop" to terminate the program.
!
! See the comments at the start of each module for its specific input
! instructions.
!
!-----------------------------------------------------------------------

NJOY is usually invoked from a terminal or command window, or from a

script. The standard I/O redirection syntax is used to control the choice of an
input file:

20 NJOY2016
2 NJOY LA-UR-17-20093

njoy < test1

Of course, the system messages that normally appear in the terminal or command
window could be redirected to a file by appending something like “ > out1” to
this execution line.

2.2 Interface Files

Another requirement of a good modular system is that the input and output
files be in a common format so that modules can work with each other’s output
in a flexible way. Since NJOY is basically an ENDF processing code, ENDF-
compatible formats were chosen for linking modules together. “Input” and “out-
put” modules can be specified to communicate with other formats (the “outside
world”). In the example above, tape23 is an example of such an external file.
The other tapes2 in the example are ENDF-type files, and the sequence shown is
fairly typical. If the user in the example needs data at a higher temperature, the
RECONR point-ENDF, or PENDF, file (tape21) can be run through BROADR
to produce a Doppler-broadened PENDF file for GROUPR. Many other combi-
nations are possible simply by rearranging the sequence of module names and
changing the unit numbers that link them. These common-format files also pro-
vide for convenient restarts at many points in the calculational sequence. For
example, if a user is trying to produce pointwise cross sections at 300K, 600K,
and 900K and runs out of time while working on 900K, he or she can save the
partially completed PENDF file and restart from 600K. Multigroup modules use
specially constructed groupwise-ENDF formats (GENDF) that are compatible
with the multigroup output modules. A GENDF file from GROUPR can be
saved in the NJOY data library, run through CCCCR to produce one output
format, and then run through MATXSR for another output format.
In NJOY, unit numbers from 20 through 99 are used for storing results or
linking modules, units 10 through 19 are reserved for scratch files, which will be
destroyed after a module has completed its job, and units below 10 are reserved
for the system. Negative unit numbers indicate binary mode.
2
The word “tape” will often be used in this report as a synonym for “file”; an actual physical tape is not
implied. This is consistent with ENDF custom, where the phrase “ENDF tape” is traditional. Furthermore,
in most NJOY installations, the actual files on the machine will have names like “tape23.”

NJOY2016 21
LA-UR-17-20093 2 NJOY

There are special utility routines to open, close, and reposition files. These
routines automatically handle the NJOY conventions on positive or negative unit
numbers, scratch files, and so on. These routines are available as the following
public calls in module util:

openz(lun,new)
Open the unit=abs(lun). If lun≥0, use coded (formatted) mode, and if
lun≤0, use binary mode. Destroy on close or job termination if 10≤lun≤20.
If new=1, destroy the file on this unit (if it exists) and open a new file.
closz(lun)
Close the file with unit=abs(lun). Do nothing if lun=0 or if lun refers to
a scratch file.
repoz(ntape)
Rewind the file with unit=abs(ntape). Do nothing if ntape=0.
skiprz(lun,nrec)
Skip nrec records forward or backwards. Caution: Some systems have a
call for this option; others can use loops of backspace and dummy reads as
given in the NJOY code. Both these operations work well for systems that
use “linked-list” data structures for I/O files. On some systems, however,
backspace is implemented as a rewind followed by forward dummy reads to
the desired location. In such cases (for example, VAX), skiprz must be
recoded to avoid calling backspace repeatedly. This caution is somewhat
moot for current operating systems.

2.3 Free Format Input

Free-form input is handled by the standard Fortran-90 READ(* method. Pre-
vious versions of NJOY used a special routine FREE for free-form input. This
routine was developed before free-form input was routinely available in Fortran,
but it has now been retired. Some capabilities that were provided by FREE,
such as repeat fields, are no longer supported using READ(*. Basically, users
can type in their input quantities separated by spaces. Lines can be terminated
early with the slash (/) symbol, leaving any variables not provided at their de-
fault values. Whether all input variables are defined in a users input record or
not we recommend that every input record be terminated with a slash (/). This
assures that all variables have been read, or assigned their default values, and
provides a more robust input file if future code changes specify additional input
variables. In the absence of a slash, fortran i/o rules may cause the next input
record to be read; an action that is certain to cause the job to crash. Text values
consisting of single words can be entered without delimiters, but more complex

22 NJOY2016
2 NJOY LA-UR-17-20093

strings containing spaces must be delimited using the single-quote character (’).
Real numbers can be entered in a variety of forms; 1, 1., 1.e0, and 1e0 should
all be equivalent. Examples of NJOY input will be found throughout this report.

2.4 ENDF Input-Output

The ENDF format for evaluated nuclear data is well documented elsewhere,[9]
but for the convenience of the reader, some features of the format will be de-
scribed here.
ENDF “tapes" are subdivided internally into “materials” (MAT), “files” (MF),
and “sections” (MT). A MAT contains all data for a particular evaluation for an
element or isotope (for example, MAT=825 is an evaluation for 16 O in ENDF/B-
VII). A file contains a particular type of data for that MAT: MF=3 is cross-
section versus energy data; MF=15 contains secondary photon energy distri-
butions. A section refers to a particular reaction [for example, MT=2 is elastic
scattering and MT=107 is the (n, α) reaction]. Every record contains the current
MAT, MF and MT values. Two materials are separated by a record with MAT=0
(the material-end or MEND record). Two files are separated by a record with
MF=0 (the file-end or FEND record). Two sections are separated by a record
with MT=0 (the section-end or SEND record). Finally, the tape is terminated
with a record with MAT = −1 (tape-end or TEND record).
NJOY has a set of utility subroutines for locating desired positions on an
ENDF tape. They are located in module endf, and they can be made available
to any other module with the statement “use endf.”

findf(mat,mf,mt,nin)
Search nin backward or forward for the first record with this MAT, MF,
MT. Issue a fatal error message if the record is not found.
tosend(nin,nout,nscr,a)
tofend(nin,nout,nscr,a)
tomend(nin,nout,nscr,a)
totend(nin,nout,nscr,a)
Skip forward past the next SEND, FEND, MEND, or TEND card on NIN.
If nout and/or nscr are nonzero, copy the records. Input and output files
must be in the same mode.

The data on an ENDF tape are written in eight different kinds of “structures”,
each of which has a binary and a formatted form. In modern systems, formatted
data is normally coded in ASCII. The words “coded,” “formatted”, and “ASCII”

NJOY2016 23
LA-UR-17-20093 2 NJOY

will often be used interchangeably here). The structures are: (1) TAPEID, a
character-string title for the tape; (2) CONT, a control record (includes SEND,
FEND, MEND, and TEND); (3) LIST, a list of data items; (4) HOLL, a list of
character-string words (what used to be called Hollerith data); (5) TAB1, a one-
dimensional tabulation of data pairs; (6) TAB2, a two-dimensional tabulation
control record; (7) INTG, a special structure of integer fields used for encoding
correlation data, and (8) DICT, a directory or index (it used to be called the
“dictionary”) to the sections found in the MAT. It should be noted that HOLL
is a special case of LIST and DICT is a special case of CONT.
In binary mode, each structure is written as a single logical record as follows:

TAPEID[MAT,MF,MT/A(I),I=1,17] 3
where MAT=tape number, MF=MT=0, and the character data are 16A4,A2;
CONT[MAT,MF,MT/C1,C2,L1,L2,N1,N2]
LIST[MAT,MF,MT/C1,C2,L1,L2,N1,N2/A(I),I=1,N1]
HOLL[MAT,MF,MT/C1,C2,L1,L2,N1,N2/A(I),I=1,N1]
where MF=1, MT=451, and each line of Hollerith characters is stored in A as
16A4,A2;
TAB1[MAT,MF,MT/C1,C2,L1,L2,N1,N2/NBT(I),JNT(I),I=1,N1
/X(I),Y(I),I=1,N2] where NBT and JNT are the interpolation table and
Y(X) is the one-dimensional tabulation;
TAB2[MAT,MF,MT/C1,C2,L1,L2,N1,N2/NBT(I),JNT(I),I=1,N1]
where the interpolation table is to be used to control a series of N2 LIST or
TAB1 structures that follow;
INTGIO[MAT,MF,MT/A(I),I=1,NW]
where each line of the array can contain 18 I3 integers, 13 I4 integers, 11
I5 integers, 9 I6 integers, or 8 I7 integers. The specific format is governed
by the value of L1 from the CONT record immediately preceding this data
structure.
DICT[MAT MF,MT/0.,0.,MFS,MTS,NCS,MODS]
where there is a record for each section in the material (MFS, MTS) giving
the card count (NCS) for that section. For ENDF/B-V, MODS indicates the
revision number for that section.

The ENDF format manual[9] explains how these structures are combined to
represent various physical quantities.
In formatted mode, each structure is broken up into many card images, each
containing 6 data words, followed by MAT, MF, MT, and a line sequence number.
3
In ENDF/B manuals, the slash is used as a logical divider. Replace it with a comma and add parentheses
when constructing a FORTRAN I/O list.

24 NJOY2016
2 NJOY LA-UR-17-20093

There is no intrinsic limit to the length of a data structure written in coded form
because a program reading the data can normally be coded to use the data in
“pages” of reasonable size. The MINX code[21] (the predecessor of NJOY) was
forced to use coded formats to handle the large tabulations found on PENDF
tapes. Analysis showed that this code used large amounts of its running time
coding and decoding number formats. In order to eliminate this waste, a blocked
binary format was developed for the ENDF data structures. A structure is
divided up into several logical records of intermediate length (typically about
300 words), each having the following form:

[MAT,MF,MT,NB,NW/A(I),I=1,NW]

where NB is the number of words remaining in the data structure (the last record
has NB=0). This type of record is compatible with the official ENDF binary
record, but is also adaptable to paging methods. The page size can be chosen to
optimize input/output rates for a particular computer system.
A set of utility subroutines has been devised to handle both blocked-binary
and paged-BCD input and output. They are also provided by module endf.

contio(nin,nout,nscr,a,nb,nw)
Read/write a control record from/to a (nb=0, nw=6). contio uses asend,
amend, etc. for END cards.
listio(nin,nout,nscr,a,nb,nw)
Read/write the first record or page of a list record from/to a. If nb is not
zero, continue with moreio, as shown in Examples 1 and 2 that follow.
tab1io(nin,nout,nscr,a,nb,nw)
Read/write the first record or page of a TAB1 structure. If nb is not zero,
use moreio.
tab2io(nin,nout,nscr,a,nb,nw)
Read/write a TAB2 structure (nb=0).
moreio(nin,nout,nscr,a,nb,nw)
Read/write continuation records or pages from/to the array a. Returns
nb=0 after processing the last record or page.
tpidio(nin,nout,nscr,a,nb,nw)
Read/write the character-string tape identification record from/to array a
(nb=0, nw=17).

NJOY2016 25
LA-UR-17-20093 2 NJOY

hdatio(nin,nout,nscr,a,nb,nw)
Read/write the first record or page of the character descriptive data (MF=1,
MT=451) from/to a, taking account of the 16A4,A2 format needed in
ASCII mode. If nb is not zero, use moreio.
intgio(nin,nout,nscr,a,nb,nw)
Read/write an INTG record.
dictio(nin,nout,nscr,a,nb,nw)
Read/write the entire material directory from/to a. On entry, nw is the
number of entries in the dictionary. moreio is not used.
asend(nout,ncr)
afend(nout,nscr)
amend(nout,nscr)
atend(nout,nscr)
Write a section (MT=0), file (MF=MT=0), material (MAT=MF=MT=0)
or tape (MAT=-1, MF=MT=0) “end” record on the desired units.

In these calling sequences, the unit numbers can be positive, negative, or

zero. Positive numbers mean ASCII mode, negative numbers mean blocked-
binary mode, and zero means the file corresponding to this position in the calling
sequence is not used. All of these routines use the following variables made public
by module endf:

c1h,c2h,l1h,l2h,n1h,n2h,math,mfh,mth,nsh,nsp,nsc

where c1h corresponds to the ENDF C1 field, and so on. The variable nsh is
the sequence number for nin, nsp is the sequence number for nout, and nsc is
the sequence number for nscr. Two examples may help to make clear the use of
these routines.

Example 1. Read All Data

loc=1
call tab1io(nin,0,0,a(loc),nb,nw)
loc=loc+nw
do while (nb.ne.0)
call moreio(nin,0,0,a(loc),nb,nw)
loc=loc+nw
enddo
(process data in A)

26 NJOY2016
2 NJOY LA-UR-17-20093

Example 2. Paging

call tab1io(nin,0,0,a(1),nb,nw)
110 (process this page of data in A)
if (nb.eq.0) go to 120
call moreio(nin,0,0,a(1),nb,nw)
go to 110
120 continue

When nin is BCD, paging is automatic. Positive and negative unit numbers can
be mixed in tpidio, contio, listio, etc., when mode conversion is desired.

2.5 Buffered Binary Scratch Storage

During the execution of a program, large amounts of data often need to be stored
in mass storage temporarily. In order to make such scratch storage as efficient
as possible, NJOY includes a pair of utility subroutines in module util that
automatically buffer such data through fast memory to disk.

loada(i,a,na,ntape,buf,nbuf)
finda(i,a,na,ntape,buf,nbuf)
Load or find data in a buffered binary scratch file. Here i is the data point
number (i must increase, except i=1 causes a rewind and i<0 flushes the
fast memory buffer to mass storage), The array a contains data to be stored
or provides a destination for data to be read, na is the number of words
to be transmitted (must be the same for all i), ntape is the logical unit
number of the disk file, buf is the fast memory buffer array, and nbuf is
the length of the buffer array.

When a point is to be saved, loada stores it in buf. When buf becomes

full, it is automatically dumped to disk. When a point is to be retrieved, finda
checks to see whether the desired point is in buf. If not, it reads through the disk
until the desired point is in memory. It then returns the desired point. When
na is small with respect to nbuf, using loada/finda reduces the number of I/O
operations dramatically. Sometimes it is necessary to find a particular part of
the buffered data. In such cases, use

scana(e,ip,np,na,ntape,buf,nbuf)

NJOY2016 27
LA-UR-17-20093 2 NJOY

where e is a value for the first of the na words, and ip points to part of the data
whose first word is either equal to e or is the first value less than e.

2.6 Dynamic Storage Allocation

Previous versions of NJOY used an internal package called storag for dynamic
memory allocation. Beginning with NJOY2012 we use the standard Fortran-90
storage allocation mechanism. The only feature used is the “allocatable array.”
When needed, an array is allocated that will contain a defined number of values.
The array can be deallocated when no longer needed, and it will disappear
automatically when the scope within which it is defined is exited. This method
produces more transparent code than the old storag package, because pointers
into a single container array are not needed.

2.7 ENDF/B Utility Routines

There are several operations performed on ENDF/B data that are needed in so
many other modules that it is practical to put them into the endf module.

terp1(x1,y1,x2,y2,x,y,i)
Interpolate for y(x) between y1 (x1 ) and y2 (x2 ) using the ENDF interpo-
lation law i (i=1 means y=y1 , i=2 means y is linear in x, i=3 means
y is linear in ln(x), i=4 means ln(y) is linear in x, i=5 means ln(y) is
linear in ln(x)), and i=6 means to interpolate using the charged-particle
penetrability with a kinematic threshold of thr6.
terpa(y,x,xnext,idis,a,ip,ir)
Interpolate for y(x) in the TAB1 structure in array a. The routine searches
for the correct interpolation range starting from ip and ir (initialize to 2
and 1 for first call). It returns xnext, the next x value in the tabulation.
idis is set to 1 if there is a discontinuity at xnext; it is zero otherwise).
gety1(x,xnext,idis,y1,itape,a)
gety2(x,xnext,idis,y1,itape,a)
Find y(x) in a TAB1 structure starting at the current location on itape
by paging the data through array a. gety1 and gety2 are identical for
occasions when two different tapes are being searched at the same time.
xnext and idis behave as in terpa. The array a must be at least npage+50
words in length. These routines are normally used to retrieve cross sections
from MF=3.
gral(xl,yl,xh,yh,x1,x2,i)

28 NJOY2016
2 NJOY LA-UR-17-20093

This function returns the integral from x1 to x2 of an ENDF function with

interpolation law i (see terp1). xl, yl, xh, and yh are the low and high
limits of the interpolation panel.
intega(f,x1,x2,a,ip,ir)
Integrate the TAB1 function stored in a from x1 to x2 . The routine auto-
matically determines the correct interpolation law for each panel or fraction
of a panel and uses gral to compute each part of the integral. Set ip=2 and
ir=1 on the first call to intega. In subsequent calls, the previous values
of ip and ir will usually provide a good starting point for searching in the
TAB1 structure.

2.8 Math Routines

Several mathematics routines are included in the mathm module for use by other
modules. Most of these routines are based on routines from the SLATEC library
converted to NJOY style and Fortran-90.

legndr(x,p,np)
Generate Legendre polynomials Pℓ (x). The ℓ = 0 value is in p(1), the
ℓ = 1 value in p(2), etc. np is the maximum Legendre order produced, so
the largest index for p is np+1.
e1(x)
Compute the first-order exponential integral function E1 (x).
gami(a,x)
Compute the incomplete gamma function γ(a, x).
erfc(x)
Compute the complementary error function erfc(x).

2.9 System-Related Utility Routines

As much as possible, actions that related to system functions have been put into
subroutines or functions in the util module. The I/O routines have already
been discussed. Sometimes these routines might have to be altered for unusual
system environments.

timer(time)
Returns the run time in seconds. The meaning of this number may vary
from system to system. It might be central-processor (CP) time, or at
some installations, it may include other factors, such as I/O time or memory
charges. This makes it difficult to compare NJOY runs on different systems.

NJOY2016 29
LA-UR-17-20093 2 NJOY

This routine may have to be revised for some systems, because there is no
standard Fortran-90 call.
dater(hdate)
Returns the date as an 8-character string in one of the forms mm/dd/yy or
ddmmyy; for example, 11/15/90 or 03jun91. For the Fortran-90 version of
NJOY, this is handled using the standard date_and_time call.
wclock(htime)
Returns the “wall clock” time. This is the time of day that the NJOY run
started, and it shouldn’t be confused with the elapsed CP time for the run
as returned by timer. The time is represented as an 8-character string in
the form hh:mm:ss; for example, 12:13:47. The user is free to use a 24-hour
convention for time. The standard date_and_time call is used to retrieve
the information from the system.
sigfig(x,ndig,idig)
Because of the many comparisons and searches that it makes, NJOY often
has to match two numbers that are different only in the few least significant
bits. This routine is intended to make such numbers exactly equal to each
other by truncating the numbers to a given number of digits and removing
any low-significance junk resulting from nonterminating binary fractions.
The idig parameter can be used to move the result up or down by idig
in the last significant figure. Although this routine was sometimes machine
dependent in previous versions of NJOY, the subroutine now used seems to
work on all systems tried so far.
a10(x,hx)
Converts x to a 10-column format as a string in hx to provide more digits
in some NJOY listings without taking too much space. This allows 4, 5, or
6 significant figures to be printed where we previously had four. hx has the
forms +1.23456+6, +1.2345-38, or -1.234+308.

2.10 Error and Warning Messages

NJOY has a pair of standard routines for printing fatal error messages and
warning messages. This helps to enforce consistency in the messages, insulates
other subroutines from the complexities of the system (for example, I/O units,
“console”, “standard error file”), and provides a site for machine-dependent error
handling, including such things as saving “drop files” and generating trace-back
listings. These routines are in module util.

error(from,mess1,mess2)
This subroutine should result in a fatal error exit and must be adjusted to
reflect the local system. Special features such as traceback information or

30 NJOY2016
2 NJOY LA-UR-17-20093

saving files for later analysis can be performed here. from is a character
string containing the name of the procedure that called error, and mess1
and mess2 are two 60-character strings containing messages describing the
error. mess2 is not printed if it is empty.
mess(from,mess1,mess2)
This routine is for nonfatal warning messages. from is the routine that
called it. It prints from, mess1, and mess2 (if not empty), and returns to
the calling routine.

The actual error messages produced by functions and subroutines in the NJOY
module and the subsidiary modules util, endf, and mathm are listed below, in-
cluding explanations of the meaning of the errors and suggested steps to alleviate
them.

error in njoy***illegal module name

Check spelling, and check for missing (/) or incorrect item counts in the pre-
ceding module. This error message is generated directly by NJOY instead
of using error.
error in openz***illegal unit number
error in closz***illegal unit number
Units less than 10 are reserved for the system.
error in tomend***mode conversion not allowed
error in tofend***mode conversion not allowed
error in tosend***mode conversion not allowed
Input and output units must both be binary or both be BCD. Check the
signs of the unit numbers in the input file.
error in findf***mat–-mf–-mt–-not on tape
Desired section cannot be found. Either the wrong tape was mounted, or
there is a mistake in the input deck.
error in scana***initial ip ne 0
Must be called with ip=0.
error in scana***did not find energy –-
Energy requested is greater than the highest energy in the LOADA/FINDA
file.
error in gral***x2 lt x1
The integration interval is bad.
error in e1***x is 0
Result is not defined.
error in gami***a must be gt zero

NJOY2016 31
LA-UR-17-20093 2 NJOY

error in gami***x must be gt zero

Illegal parameter values.
error in dlngam***abs(x) so big dlngam overflows
error in dlngam***x is a negative integer
Illegal parameter values.
message from dlngam–-answer lt half precision...
Just a warning.
error in csevl***number if terms le 0
error in csevl***number if terms gt 1000
Won’t happen in NJOY.
error in csevl***x outside the interval (-1,1)
Bad parameter value.
error in d9lgit***x should be gt 0 and le a
error in d9lgit***no convergence in 200 terms...
Problems computing Tricomi’s incomplete gamma function. This error
should not occur.
message from d9lgit–-result less than half precision...
Just a warning.
error in d9lgmc***x must be ge 10
Bad parameter value.
error in dgamit***x is negative
Bad parameter value.
message from dgamit–-result less than half precision...
Just a warning.
error in dgamlm***unable to find xmin
error in dgamlm***unable to find xmax
Having trouble finding the minimum and maximum bounds for the argu-
ment in the gamma function.
error in dgamma***x is 0
error in dgamma***x is a negative integer
error in dgamma***x so big gamma overflows
Bad parameters for the gamma function.
message from dgamma–-answer lt half precision...
Just a warning.
error in d9gmit***x should be gt 0
error in d9gmit***no convergence in 200 terms...
Problem’s with Tricomi’s gamma function at small arguments.
error in d9lgic***no convergence in 300 terms...
Problems with the log complementary incomplete gamma function.

32 NJOY2016
2 NJOY LA-UR-17-20093

2.11 Coding Details for the NJOY Main Program

The njoy main program of the NJOY code starts with a with a block of comment
cards that gives a short description of the NJOY module and specifications for
the user’s input lines. (The term “card” is used out of respect for the past;
this usage should not be taken to imply that a real card that can be “folded,
spindled, or mutilated” has to be used.) These blocks of comments cards occur
at the beginning of every NJOY module. It is a good idea to check the input
instructions in the comment cards for the current version in order to see whether
there have been any changes from the input instructions reproduced in this
manual.
The first step in the body of the code is to give “use module” statements for
each module that provides a variable or subroutine call to be used here. There
are several common variables used in this module. The vers string and its
associated date string vday from the version module are important parts of the
NJOY Quality Assurance (QA) scheme. They are updated each time a change
is made to the code, and they are always printed on the output listing. They are
also available to other modules to be written in libraries generated by NJOY or
on plots, if desired. The lab and mx strings are normally used to localize NJOY
to a particular institution and to tell what machine was used for an NJOY run
when several are available. They are carried in the locale module. The quantity
npage in the endf module must be a multiple of 6 (because ENDF records have 6
fields on a line) and 17 (since Hollerith lines use the format “16a4,a2”); therefore,
a value of 306 was chosen. We do not recommend changing this value.
Next, NJOY writes an “banner” on the output listing file giving the date,
time, version, and so on, for this NJOY run. The program now starts an infinite
loop, reading in module names, and executing the requested module, until the
name “stop” is read.

Module “locale” This purpose for this module is to provide localization infor-
mation for NJOY. Users may find it necessary to change things in this module
for their site, machines, or compilers. The public variables provided are

lab – a string identifying the user’s institution.

mx – a string identifying the machine used, the system, the compiler, or what-
ever.
kr – the kind value for normal real numbers.

NJOY2016 33
LA-UR-17-20093 2 NJOY

k4 – the kind value for 4-byte real numbers and 4-byte integers (mostly for
CCCC records).
k8 – the kind value for 8-byte Hollerith variables in the CCCC formats.

The default value for lab is “our-lab,” and it can be changed to reflect the
users site. Up to 8 characters are allowed. The default value for mx is “our-mx.”
At one time, LANL had a system call that would tell which machine out of a
cluster was being used, and this field was used for that. It can also be used to
identify the type of machine (e.g., sun, vax, x86, linux) or the compiler used.
NJOY2016 uses the built-in features of Fortran-90 to control the precision of
the numbers used internally. This allows us to remove the complex short-word
and long-word controls used for the Fortran-77 versions, a great simplification.
The normal internal representation used in NJOY2016 is a high-precision one,
normally implemented using 64-bit reals and the system default for integers
(either 4 or 8 bytes should be OK). The “kind” value to obtain these high-
precision reals is returned using the Fortran-90 selected_real_kind function,
and it should be portable. However, k4 and k8 are simply set to the values 4
and 8, respectively. These choices are not standardized by Fortran-90, and the
values given here may have to be adjusted for some systems. Anyway, once a
proper value of kr has been provided, variables, arrays, and constants can be
typed using statements like

real(kr)::za
real(kr)::elast(20)
real(kr),dimension(:),allocatable::enode
real(kr),parameter::therm=.0253e0_kr

Module “version” This module provides public version and date strings to
any of the other NJOY modules. The initial values are vers=’2016.0’ and
vday=’xxdec16’. It is an important part of the NJOY QA procedure to keep
these values up to date as changes are made to the code.

Module “mainio” This module provides public values for the system I/O
units, namely, nsysi, nsyso, and nsyse. In the past we assigned fixed numeric
values to these variables based upon legacy fortran definitions. We now use the
iso_fortran_env package to define these units, thereby providing greater porta-
bility over multiple computing platforms. Specifically, nsysi, used in “READ(nsysi,...”

34 NJOY2016
2 NJOY LA-UR-17-20093

statements was set equal to 5 in the past and is now set to INPUT_UNIT. nsyso,
used in “WRITE(nsyso,...” statements was set equal to 7 previously and is now
given by ERROR_UNIT. Finally, nsyse, used in “WRITE(nsyse,...” statements
directed to the terminal, was set equal to 6 previously and is now given by
OUTPUT_UNIT. As used by NJOY the “ERROR_UNIT” and “OUTPUT_UNIT”
names are counter-intuitive since write(nsyse, ...) goes to the terminal while
write(nsyso, ...) goes to a file called “output”, a file that is opened in the main
NJOY program. As noted elsewhere, NJOY uses unit numbers between 10 and
19 for scratch files, and allows users to specify unit numbers from 20 and above
for file i/o between NJOY modules. If users encounter conflicts on their local
platforms with the values in iso_fortran_env we recommend they revert to the
legacy definitions noted above.

Module “physics” Because NJOY is divided into a number of separate mod-

ules, it is important to provide a common set of physical constants to ensure
consistency. These constants are as follows:

pi – a high precision of 15 digits is provided.

bk – Boltzmann’s constant.
amassn – the mass of the neutron in AMU.
amu – the value of the AMU.
hbar – Planck’s constant divided by 2π.
ev – the value of the electron volt.
clight – the velocity of light.

These numbers were obtained from the CODATA’89 set as published by the
National Institute of Standards and Technology (NIST). Other NJOY modules
can access these values by including the statement “use physics.” We have
taken pains in NJOY to compute values that can be derived from these standards
from these values rather than multiplying the number of physical constants used
in the program.

Module “util” As outlined above, module util makes the following routines
public:

error – fatal error routine

mess – warning message routine

NJOY2016 35
LA-UR-17-20093 2 NJOY

timer – CPU elapsed time routine

dater – returns a date string
wclock – returns a wall-clock time string
repoz – repositions a file to the beginning
skiprz – skips records in a file forward and back
openz – opens a new or exiting file
closz – closes a file
loada/finda – a buffered storage system
scana – search in a loada/finda system
sigfig – truncate numbers to a given significance
a10 – print real numbers compactly

There are no private subroutines or variables in this module.

The default versions of error and mess simply write the one or two lines of
message characters to the appropriate units. The Fortran-90 len_trim function
is used to measure the real length of a character string by removing trailing
blanks. By changing nsyse, messages can be directed to the user’s terminal, to
a standard error unit, or left to appear on the output file only. The default fatal
error exit “stop 77” can be sometimes be changed to cause a “traceback” to the
subroutine that called error, and backward through the stack of subroutines
that called it. Also, some systems can cause a drop file or “core” file to be
generated for post-mortem analysis with a debugging program.
Subroutines timer, dater, and wclock provide standard interfaces to system
routines that sometimes have different names and return different kinds of an-
swers on different computer systems. This is less of a problem with Fortran-90
than it was with previous compilers. The date and wall-clock strings are easy to
construct using the standard date_and_time call, which is portable to all sys-
tems. Some system have an analogous cpu_time call, but it is not standard. For
other systems, it is necessary to use machine-dependent calls. Some examples of
implementing this function on various computer systems are included as UPD
machine-dependent idents in the NJOY distribution.
The next four routines in NJOY provide a uniform way of handling input and
output files. The unit numbers obey the following convention: zero means do
nothing, negative means a binary unit, and positive means a coded unit (ASCII
for modern systems.). Unit numbers between 10 and 19 are scratch files, which
will be automatically destroyed at the end of the job. Therefore, repoz simply
takes the absolute value of the unit number and calls rewind if it is nonzero.

36 NJOY2016
2 NJOY LA-UR-17-20093

Similarly, closz takes the absolute value of the unit number and closes it if
nonzero and not a scratch file. Subroutine skiprz skips forward with dummy
reads of the correct type for the sign of the unit, and it skips backward with
backspace. Note that this can be very inefficient for some systems. For example,
the VAX executes backspace by first rewinding and then skipping forward by
n−1 records. Making successive calls to backspace, as done in this routine,
would be very expensive. It is better to rewind before calling skiprz; then only
forward skips would be required to get to the desired record. Finally, subroutine
openz is used to open files with the desired characteristics (that is, binary or
formatted, new or old, scratch or permanent). Standard Fortran-90 statements
are used. Note that the file names are constructed to have the forms tape20,
tape21, etc. Scratch files will have whatever names are standard for the system
being used.
The loada/finda system was discussed in Section2.5. The data representa-
tion consists of a fairly large buffer array buf of length nbuf. This buffer contains
a number of component blocks, each of length na. Therefore, it is easy to com-
pute the location of any block using modular arithmetic on the block index i.
If the location is currently in the buffer array, the block can be read or written.
Otherwise, the associated scratch file ntape must be repositioned before being
read or written. The files ntape have to be written sequentially, but they can be
searched in any order. Subroutine scana takes advantage of this to locate the
block whose first element is closest to the input parameter e.
Subroutine sigfig is used to control the precision of numbers in several NJOY
modules. This is a difficult problem because the true representation of numbers
in the machine is as non-terminating binary fractions. In addition, different
representations for floating-point numbers are used in different computer sys-
tems, and significant-figure truncation can be machine-dependent. The version
of sigfig used in NJOY2016 seems to be very portable.
Subroutine a10 is used to make the printing of real numbers on NJOY output
listings more compact by stripping off excess characters in the exponent field.
It converts the input real number into a 10-character string with the following
forms: +1.23456+6, +1.2345-38 and -1.234+308. This subroutine is closely
related to the private subroutine a11 in the endf module, except for forcing a
10-character field rather than an 11-character field.

Module “endf ” As discussed in Section 2.4 of this chapter, NJOY uses files
in ENDF-like formats for most communications between modules. Subroutines

NJOY2016 37
LA-UR-17-20093 2 NJOY

are provided in module endf to work with the various ENDF record types. The
routines that are public for this module are as follows:

contio – to read or write CONT records

listio – to read or write LIST records
tab1io – to read or write TAB1 records
moreio – to read or write continuation records
tab2io – to read or write TAB2 records
tpidio – to read or write TAPEID records
hdatio – to read or write character-string records
intgio – to read or write INTG reconrds
dictio – to read or write “dictionary” records
tosend – to read or write to the next SEND record
tofend – to read or write to the next FEND record
tomend – to read or write to the next MEND record
totend – to read or write to the next TEND record
asend – to write SEND record
afend – to write FEND record
amend – to write MEND record
atend – to write TEND record

In addition, the following data words are public for this module:

c1h – ENDF C1 field

c2h – ENDF C2 field
l1h – ENDF L1 field
l2h – ENDF L2 field
n1h – ENDF N1 field
n2h – ENDF N2 field
math – ENDF MAT field
mf – ENDF MF field
mt – ENDF MT field
nsh – NS card number (input file)
nsp – NS card number (output file)
nsc – NS card number (scratch file)
thr6 – threshold energy for interpolation law 6

38 NJOY2016
2 NJOY LA-UR-17-20093

There are additional private routines and global variables used inside the module.
Subroutines contio, listio, tab1io, tab2io, moreio, tpidio, hdatio, intgio,
and dictio are used to read, write, copy, or convert the mode of binary or for-
matted records on these interface files. They all have similar structures. First,
records are read using binary or coded commands, depending on the sign of
the unit number. Then the records are written back out, once again using a
mode that depends on the sign of the unit number. Unit numbers can be zero,
in which case that unit is not used. When these numbers are read, they are
converted into integers for the l1h, l2h, n1h, and n2h fields using the Fortran
nearest-integer function, nint. If programmers choose to use a number in the
array a directly, they should be careful to make the same conversion. Private
subroutine lineio is used by several of these routines to read or write a line of
floating-point numbers. It has two special features. First, any empty fields at
the end of a line are filled in with blanks. The second feature is the construc-
tion of floating-point numbers without the normal Fortran “E” using subroutine
a11. Private subroutine tablio is used by tab1io and tab2io to read and write
ENDF-style interpolation tables. Subroutine hollio is used to read or write
lines of Hollerith information. Note that Hollerith data are represented using 17
fields per card, but all other ENDF data uses 6 fields per card.
Subroutine a11 simply breaks up the real-number x into a fraction part f,
a sign part S, and an exponent part n. The number of digits in the fraction
part depends on the number of digits in the exponent part in order to maximize
the precision of the formatted representation of x. It then constructs a string
containing the real number without the normal Fortran “E” and omitting leading
zeros in the exponent. If the number being represented has more significant digits
than can be represented with the “E” formats, the subroutine automatically shifts
to an appropriate “F” format (e.g., F11.8, F11.7, |it etc.) depending on the size
of the number. This feature is sometimes needed for the fine texture of resonance
data.
As discussed above, ENDF tapes are divided into materials, files, and sections
by special end cards with names like MEND, FEND, SEND, and TEND. The
subroutines tosend, tofend, tomend, and totend can be used to move from the
current location to one of these end cards, optionally copying the records as they
go. Since these routines do not know the actual structure of the records, they
cannot change the mode of the data while copying. They simply read in a record,
write it out again if desired, and watch for the specified end card.
The parallel set of routines asend, afend, amend, and atend write one of the

NJOY2016 39
LA-UR-17-20093 2 NJOY

requested end cards to one or two different output files using the mode defined
by the sign of the unit number. The NJOY convention is to use blanks for the
first 66 columns of formatted end cards in order to make them easy to see on
ENDF listings. The tapes as received from the National Nuclear Data Center
(NNDC) of the Brookhaven National Laboratory may have these fields filled
in with numerical values of zero. The MODER module automatically changes
ENDF tapes from the BNL convention to the NJOY one.
Subroutine findf is used very frequently in NJOY modules to search through
an ENDF-type tape for a desired section (mat, mf, mt). Since the three numbers
that describe sections are always arranged in order, it is possible to search both
up and down. On entry, the routine reads the first card and decides if it has to
read up or down to find the desired section. It then continues by reading records
in the proper direction until it comes to the desired section. It then backs up
by one record so that the next read after the call to findf will read the first
record of the desired section. If the desired section is not found, a fatal error
message will be issued. Subroutine findf can go into an infinite loop if the MAT
number requested is smaller than the number for the first material on the tape
(thus findf is moving backwards) and the number in the MAT field on the tape
identification record is larger than the material number for the first material.
Some systems have an “IF (BOI)” function that can be used to detect when the
tape is at the “beginning of information.”
Subroutine terp1 is used to interpolate between two points x1,y1 and x2,y2
using ENDF interpolation law i. The results for y at x are given by simple for-
mulas. No tests are made for values that give illegal arguments for the Fortran
log and exponent functions. Therefore, fatal arithmetic errors from terp1 are
fairly common. Going slightly out of order, gral is a routine that computes
integrals inside the endpoints of a segment represented using the ENDF interpo-
lation laws. Once again, simple formulas are used, and it is also possible to get
arithmetic errors from the log and exponent functions with this routine.
It is usually necessary to interpolate or integrate functions that involve more
than one panel of ENDF interpolated data. Most commonly, this occurs for
data in TAB1 format. If the TAB1 record is small enough so that it can fit into
an array a in memory, subroutine terpa can be used to search for the panel
that contains x, and to interpolate for y at x. Similarly, intega can be used to
compute the integral from x1 to x2 by integrating over all the panels and parts of
panels within these limits. Subroutine terpa uses terp1, and subroutine intega
uses gral.

40 NJOY2016
2 NJOY LA-UR-17-20093

NJOY works with some very large TAB1 records that cannot fit into memory.
For these cases, gety1 and gety2 can be used to interpolate for y at x. Array
a only has to be big enough for one page of data (plus a little extra space for
parameters like c1h, n2h, and TAB1 interpolation tables). If the desired x is
not in memory, new pages of data are read from itape until the desired value is
found. The subroutine then uses terp1 to compute y. Note that zero is returned
if x is outside the range of the table, and xnext=1.e12 when the last point is
returned. Subroutines gety1 and gety2 are almost identical so that numbers
can be retrieved independently from two different input tapes at the same time.

Module “mathm” There are several math routines that are needed in various
NJOY modules that are provided in this module:

legndr – provides Legendre polynomials

e1 – provides the first-order exponential integral function.
gami – provides the incomplete gamma function.
erfc – provides the complementary error function.

The last three of these were taken from the SLATEC library, which was developed
by the Los Alamos National Laboratory, the Sandia National Laboratory, and
the Stanford Linear Accelerator Laboratory. These routines produce very high-
precision results as compared to other publicly available routines. They were
converted to NJOY conventions and Fortran-90 style for NJOY. This module
also contains a large number of private routines that are used in the course of
computing the numbers returned by the public routines. Subroutine legndr is
used to compute Legendre polynomials by recursion. The formulas are as follows:

P0 (x) = 1. , (1)
P1 (x) = x , (2)
(2ℓ − 1)xPℓ (x) − (ℓ − 1)Pℓ−2
Pℓ+1) (x) = , (3)
ℓ

where P0 is stored in p(1), P1 is stored in p(2), and so on.

NJOY2016 41
LA-UR-17-20093 2 NJOY

42 NJOY2016
3 RECONR LA-UR-17-20093

3 RECONR
The RECONR module is used to reconstruct resonance cross sections from res-
onance parameters and to reconstruct cross sections from ENDF nonlinear in-
terpolation schemes. The output is written as a pointwise-ENDF (PENDF) file
with all cross sections on a unionized energy grid suitable for linear interpola-
tion to within a specified tolerance. Redundant reactions (for example, total
inelastic, charged-particle reactions) are reconstructed to be exactly equal to the
sum of their reconstructed and linearized parts at all energies. The resonance
parameters are removed from File 2, and the material directory is corrected to
reflect all changes. RECONR has the following features:

• Efficient use of dynamic storage allocation and a special stack structure

allow very large problems to be run.
• The unionized grid improves the accuracy, usefulness, and ENDF compat-
ibility of the output. All summation cross sections are preserved on the
union grid. Up to nine significant figures are allowed.
• A correct directory of the output tape is provided.
• Approximate ψχ Doppler broadening may be used in some cases to speed
up reconstruction.
• A resonance-integral criterion is added to the normal linearization criterion
in order to reduce the number of points added to the tabulation to represent
“unimportant” resonances.
• All ENDF-6 resonance formats currently active are handled, including the
calculation of angular distributions from resonance parameters in some
cases.

This chapter describes the RECONR module in NJOY2016.0.

3.1 ENDF/B Cross Section Representations

A typical cross section derived from an ENDF/B evaluation is shown in Fig. 1.
The low-energy cross sections are “smooth”. They are described in File 3 (see
Section 2.4 for a review of ENDF/B nomenclature) using cross-section values
given on an energy grid with a specified law for interpolation between the points.
In the resolved resonance range, resonance parameters are given in File 2, and
the cross sections for resonance reactions have to be obtained by adding the
contributions of all the resonances to “backgrounds” from File 3. At still higher
energies comes the unresolved region where explicit resonances are no longer
defined. Instead, the cross section is computed from statistical distributions

NJOY2016 43
LA-UR-17-20093 3 RECONR

of the resonance parameters given in File 2 and backgrounds from File 3 (or
optionally taken directly from File 3 as for smooth cross sections). Finally, at
the highest energies, the smooth File 3 representation is used again.
For light and medium-mass isotopes, the unresolved range is usually omit-
ted. For the lightest isotopes, the resolved range is also omitted, the resonance
cross sections being given directly in the “smooth” format. In addition, sev-
eral different resonance representations are supported (Single-Level Breit-Wigner
(SLBW), Multilevel Breit-Wigner (MLBW), Adler-Adler, Hybrid R-Function
(HRF), Reich-Moore (RM), Reich-Moore-Limited (RML), energy-independent
unresolved, and energy-dependent unresolved). The Adler-Adler and Hybrid
formats are not being used in modern evaluations. For an increasing number of
modern evaluations, the low energy “smooth” region is omitted, and the resolved
resonance region is extended to the low energy limit.
RECONR takes these separate representations and produces a simple cross
section versus energy representation like the one shown in Fig. 1.

Figure 1: A typical cross section reconstructed from an ENDF/B evaluation using RE-
CONR. The smooth, resolved, and unresolved energy regions use different rep-
resentations of the cross sections. This is the total cross section for 235 U from
ENDF/B-V.

44 NJOY2016
3 RECONR LA-UR-17-20093

3.2 Unionization and Linearization Strategy

Several of the cross sections found in ENDF/B evaluations are summation cross
sections (for example, total, inelastic, sometimes (n,2n) or fission, and some-
times charged-particle reactions), and it is important that each summation cross
section be equal to the sum of its parts. However, if the partial cross sections
are represented with nonlinear interpolation schemes, the sum cannot be rep-
resented by any simple interpolation law. A typical case is the sum of elastic
scattering (MT=2 interpolated linearly to represent a constant) and radiative
capture (MT=102 interpolated log-log to represent 1/v). The total cross section
cannot be represented accurately by either scheme unless the grid points are very
close together. This effect leads to significant balance errors in multigroup trans-
port codes and to splitting problems in continuous-energy Monte Carlo codes.
The use of linear-linear interpolation (i.e., σ linear in E) can be advanta-
geous in several ways. The data can be plotted easily, they can be integrated
easily, cross sections can be Doppler broadened efficiently (see BROADR), and,
linear data can be retrieved efficiently in continuous-energy Monte Carlo codes.
Therefore, RECONR puts all cross sections on a single unionized grid suitable
for linear interpolation. As described in more detail below, RECONR makes one
pass through the ENDF/B material to select the energy grid, and then a second
pass to compute cross sections on this grid. Each cross section on the PENDF
file (except for the summation cross sections) is exactly equal to its ENDF/B
value. The summation cross sections are then obtained by adding up the partial
cross sections at each grid point.
While RECONR is going through the reactions given in the ENDF/B eval-
uation, it also checks the reaction thresholds against the Q value and atomic
weight ratio to the neutron A (AWR in the file) given for the reaction. If

A+1
threshold ≥ Q (4)
A
is not true, the threshold energy is moved up to satisfy the condition. This is usu-
ally a small change, often only in the least significant digit, and is a consequence
of comparing two REAL numbers of finite precision.
If desired, the unionized grid developed from the ENDF/B file can be supple-
mented with “user grid points” given in the input data. The code automatically
adds the conventional thermal point of 0.0253 eV and the 1, 2, and 5 points in
each decade to the grid if they are not already present. These simple energy grid
points help when comparing materials, and they provide well-controlled starting

NJOY2016 45
LA-UR-17-20093 3 RECONR

points for further subdivision of the energy grid.

There are special problems with choosing the energy grid in the unresolved
range. In some cases, the unresolved cross section is represented using resonance
parameters that are independent of energy. The cross sections are not constant,
however, but have a shape determined by the energy variation of neutron wave
number, penetrability factors, and so on. RECONR handles this case by choosing
a set of energies (about 13 per decade) to be used to calculate the cross sections;
the set of energies gives a reasonable approximation to the result intended. For
evaluations that use energy-dependent resonance parameters, it is supposed to
be sufficient to compute the unresolved cross sections at the given energies and
to use interpolation on the cross sections to obtain the appropriate values at
other energies. However, some evaluations carried over from earlier versions of
ENDF/B were not evaluated using this convention, and cross sections computed
using cross-section interpolation are not sufficiently accurate. Even some modern
evaluations use inadequate energy grids for the unresolved range. RECONR
detects such cases by looking for large steps between the points of the given
energy grid. It then adds additional energy grid points using the same 13-per-
decade rule used for energy-independent parameters. “Large” is currently defined
by wide to be a factor of 1.26.

3.3 Linearization and Reconstruction Methods

Linearization (lunion) and resonance reconstruction (resxs) both function by
inserting new energy grid points between the points of an original grid using
an “inverted stack”. The general concepts involved are illustrated with a simple
example shown in Fig. 2.
The stack is first primed with two starting values. For linearization, they will
be two adjacent points on the original union grid. For reconstruction, they will
usually be the peaks and half-height energies of resonances. The stack is said to
be inverted because the lower energy is at the “top” (I=2).
This interval or panel is now divided into two parts, and the cross section
computed at the intermediate point is compared to the result of linear interpo-
lation between the adjacent points. If the two values do not agree within various
criteria, the top of the stack is moved up one notch (I=3), and the new value is
inserted (I=2). The code then repeats the checking process for the new (smaller)
interval at the top of the stack. The top of the stack rises until convergence is
achieved for the top interval. The top energy and cross section are then saved on
a scratch file, the stack index is decremented, and the checks are repeated. This

46 NJOY2016
3 RECONR LA-UR-17-20093

cross
section

energy

2 1
1.
3 2 1
2.
3. 4 3 2 1
5 4 3 2 1
4.
4 3 2 1
5.
3 2 1
6.
4 3 2 1
7.
3 2 1
8.
2 1
9.

Figure 2: Inverted -stack method used in RECONR and several other places in NJOY. Line
1 shows the two initial points (the lower energy is higher in the stack). In line 2, a
new point has been calculated at the midpoint, but the result was not converged,
and the new point has been inserted in the stack. In line 3, the midpoint of the
top panel has been checked again, found to be not converged, and inserted into
the stack. The same thing happens in line 4. In line 5, the top panel is found to be
converged, and the top point (5) has been written out. The same thing happens
in line 6. In line 7, the top panel is tested and found to be not converged. The
midpoint is added to the stack. Finally, in line 8, the top panel is found to be
converged, and the top point is written out. This leaves two points in the stack
(see line 9). Note that the energy points come off the stack in the desired order
of increasing energy, and that only one point has to be moved up in the stack as
each new result is inserted.

NJOY2016 47
LA-UR-17-20093 3 RECONR

process is continued with the top of the stack rising and falling in response to
the complexity of the cross section until the entire panel ∆E has been converged
(I=1). The stack is then reprimed with the bounds of the next panel. The pro-
cess continues until the entire energy range for linearization or reconstruction
has been processed.
This stack logic enables a panel to be subdivided into parts as small as ∆E/2n
where n is the stack size, and several different cross sections (elastic, capture,
fission) can easily be stored in arrays of this size.
The convergence criterion used for linearization is that the linearized cross
section at the intermediate point is within the fractional tolerance err (or a
small absolute value errlim) of the actual cross section specified by the ENDF
law. More complicated criteria are used for resonance reconstruction.
There are two basic problems that arise if a simple fractional tolerance test is
used to control resonance reconstruction. First, as points are added to the energy
grid, adjacent energy values may become so close that they will be rounded to
the same number when a formatted output file is produced. There can be seri-
ous problems if the code continues to add grid points after this limit is reached.
Through the use of dynamic format reconstruction, the energy resolution avail-
able for formatted NJOY output (which use ENDF 11-character fields) is 7 sig-
nificant figures (that is, ±1.234567 ± n) rather than the usual 5 or 6 (see Section
2.4). For NJOY2016, the Fortran-90 “kind” parameter is used to assure sufficient
precision for this. Even this seven significant figure format is sometimes insuffi-
cient for very narrow resonances. If necessary, NJOY can go to nine significant
figures by using a Fortran “F” format, e.g., ±1234.56789.
Significant-figure control is implemented as follows: each intermediate energy
is first truncated to 7 significant figures before the corresponding cross sections
are computed. If the resulting number is equal to either of the adjacent values
and convergence has not been obtained, subdivision continues using energies
truncated to 9 significant figures. If an energy on this finer grid is equal to either
of the adjacent values, the interval is declared to be converged even though
convergence has not been achieved. Thus, no identical energies are produced,
but an unpredictable but very small loss in accuracy results.
The second basic problem alluded to above is that a very large number of
resonance grid points arise from straightforward linear reconstruction of the res-
onance cross section of some isotopes. Many of these points come from narrow,
weak, high-energy resonances, which do not need to be treated accurately in
many applications. As an example, the capture and fission resonance integrals

48 NJOY2016
3 RECONR LA-UR-17-20093

important for thermal reactors must be computed with a 1/E flux weighting. If
the resonance reconstruction tolerance is set high (say 1%) to reduce the cost of
processing, the resonance integrals will be computed to only 1% accuracy. How-
ever, if the reconstruction tolerance were set to a smaller value, like 0.1%, and if
the high-energy resonances (whose importance is reduced by the 1/E weight and
the 1/v trend of the capture and fission cross sections) were treated with less
accuracy than the low-energy resonances, then it is likely that one could achieve
an accuracy much better than 1% with an overall reduction in the number of
points (hence computing cost). Since 1/E weighting is not realistic in all appli-
cations (for example, in fast reactors), user control of this “thinning” operation
must be provided.
Based on these arguments, the following approach was chosen to control the
problem of very large files. First, panels are subdivided until the elastic, capture,
and fission cross sections are converged to within errmax, where errmax ≥ err.
These two tolerances are normally chosen to form a reasonable band, such as 1%
and 0.1%, to ensure that all resonances are treated at least roughly (for example,
for plotting). If the resonance integral (1/E weight) in some panel is large, the
panel is further subdivided to achieve an accuracy of err (say 0.1%). However,
if the contribution to the resonance integral from any one interval gets small, the
interval will be declared converged, and the local value of the cross section will
end up with some intermediate accuracy. The contribution to the error in the
resonance integral should be less than 0.5×∆σ×∆E. This value is added into
an accumulating estimate of the error, and a count of panels truncated by the
resonance integral check is incremented.
The problem with this test is that RECONR does not know the value of the
resonance integral in advance, so the tolerance parameter errint is not the actual
allowed fractional error in the integral. Instead, it is more like the resonance
integral error per grid point (barns/point). Thus, a choice of errint=err/10000
with err=0.001 would limit the integral error to about 0.001 barn if 10000 points
resulted from reconstruction. Since important resonance integrals vary from a
few barns to a few hundred barns, this is a reasonable choice. The integral check
can be suppressed by setting errint very small or errmax=err.
When resonance reconstruction is complete, RECONR provides a summary
of the possible resonance integral error due to the integral check over several
coarse energy bands. An example from ENDF/B-VII.0 235 U follows:

NJOY2016 49
LA-UR-17-20093 3 RECONR

estimated maximum error due to

resonance integral check (errmax,errint)

upper elastic percent capture percent fission percent

energy integral error integral error integral error
1.00E-05
1.00E-04 3.50E+01 0.000 8.15E+03 0.000 4.29E+04 0.000
1.00E-03 3.50E+01 0.000 2.57E+03 0.000 1.36E+04 0.000
1.00E-02 3.50E+01 0.000 8.02E+02 0.000 4.24E+03 0.000
1.00E-01 3.46E+01 0.000 2.15E+02 0.000 1.26E+03 0.000
1.00E+00 3.25E+01 0.000 6.03E+01 0.000 3.26E+02 0.000
2.00E+00 8.95E+00 0.000 7.31E+00 0.000 2.62E+01 0.000
5.00E+00 1.08E+01 0.000 1.25E+01 0.000 1.56E+01 0.000
1.00E+01 7.75E+00 0.000 2.40E+01 0.000 3.52E+01 0.000
2.00E+01 8.18E+00 0.000 2.92E+01 0.000 3.31E+01 0.000
5.00E+01 1.07E+01 0.000 2.57E+01 0.000 3.83E+01 0.000
1.00E+02 8.30E+00 0.000 1.07E+01 0.000 2.34E+01 0.000
2.00E+02 8.04E+00 0.000 8.17E+00 0.000 1.42E+01 0.000
5.00E+02 1.10E+01 0.001 6.81E+00 0.008 1.51E+01 0.004
1.00E+03 8.28E+00 0.008 3.44E+00 0.080 7.62E+00 0.038
2.00E+03 8.27E+00 0.033 2.54E+00 0.261 5.06E+00 0.185

points added by resonance reconstruction = 232418

points affected by resonance integral check = 80445
final number of resonance points = 242170
number of points in final unionized grid = 242600

The parameters errmax and errint, taken together, should be considered as ad-
justment “knobs” that can increase or decrease the errors in the “res-int” columns
to get an appropriate balance between accuracy and economy for a particular ap-
plication. The error from significant figure reduction provided by earlier versions
of NJOY is no longer needed.
For energies in the thermal range (energies less than trange=0.5 eV), the
user’s reconstruction tolerance is divided by a factor of 5 in order to give better
results for several important thermal integrals, especially after Doppler broaden-
ing, and to make the 0.0253 eV cross section behave well under Doppler broad-
ening.

50 NJOY2016
3 RECONR LA-UR-17-20093

3.4 Resonance Representations

RECONR uses the resonance formulas as implemented in the original RESEND
code[26] with several changes: a more efficient calculation of MLBW cross sec-
tions developed by C. R. Lubitz of the Knolls Atomic Power Laboratory (KAPL)
and coded by P. Rose of the National Nuclear Data Center (NNDC) at the
Brookhaven National Laboratory (BNL), the addition of competitive widths
introduced for ENDF/B-V, a ψχ Doppler-broadening calculation for SLBW
and Adler-Adler resonance shapes, and a capability to process either the multi-
level multi-channel R-matrix Reich-Moore parameters or the multi-level single-
channel Hybrid R-Function parameters based on the work of M. Bhat and C.
Dunford of the NNDC, an implementation of the GH method for MLBW res-
onances, which allows psi-chi broadening, and a capability to process the new
RML parameters, including resolved resonance energy region angular distribu-
tions.
Expanded discussions of the following formulas can be found in the ENDF-6
format manual[9].

Single-Level Breit-Wigner Representation (SLBW) The subroutine that

computes Single-Level Breit-Wigner cross sections (csslbw) uses

σn = σp
XX nh Γnr i
+ σmr cos 2ϕℓ − (1 − ) ψ(θ, x)
r
Γr
ℓ
o
+ sin 2ϕℓ χ(θ, x) , (5)
XX Γf r
σf = σmr ψ(θ, x) , (6)
r
Γ r
ℓ
XX Γγr
σγ = σmr ψ(θ, x) , and (7)
r
Γr
ℓ
X 4π
σp = (2ℓ + 1) sin2 θℓ , (8)
k2
ℓ

where σn , σf , σγ , and σp are the neutron (elastic), fission, radiative capture, and
potential scattering components of the cross section arising from the given reso-
nances. There can be “background” cross sections in File 3 that must be added
to these values to account for competitive reactions such as inelastic scattering
or to correct for the inadequacies of the single-level representation with regard
to multilevel effects or missed resonances. The sums extend over all the ℓ values

NJOY2016 51
LA-UR-17-20093 3 RECONR

and all the resolved resonances r with a particular value of ℓ. Each resonance
is characterized by its total, neutron, fission, and capture widths (Γ, Γn , Γf , Γγ ),
by its J value (AJ in the file), and by its maximum value (smax= σmr /Γr in the
code)
4π Γnr
σmr = 2
gJ , (9)
k Γr
where gJ is the spin statistical factor

2J + 1
gJ = , (10)
4I + 2

and I is the total spin (SPI) given in File 2, and k is the neutron wave number,
which depends on incident energy E and the atomic weight ratio to the neutron
for the isotope A (AWRI in the file), as follows:

A √
k = (2.196771×10−3 ) E. (11)
A+1

There are two different characteristic lengths that appear in the ENDF resonance
formulas: first, there is the “scattering radius” â, which is given directly in File
2 as AP; and second, there is the “channel radius” a, which is given by

a = 0.123 A1/3 + 0.08 . (12)

If the File 2 parameter NAPS is equal to one, a is set equal to â in calculating

penetrabilities and shift factors (see below). The ENDF-6 option to enter an
energy-dependent scattering radius is not supported. The neutron width in the
equations for the SLBW cross sections is energy dependent due to the penetration
factors Pℓ ; that is,
Pℓ (E) Γnr
Γnr (E) = , (13)
Pℓ (|Er |)
where

P0 = ρ , (14)
ρ3
P1 = , (15)
1 + ρ2
ρ5
P2 = , (16)
9 + 3ρ2 + ρ4
ρ7
P3 = , and (17)
225 + 45ρ2 + 6ρ4 + ρ6
ρ9
P4 = , (18)
11025 + 1575ρ2 + 135ρ4 + 10ρ6 + ρ8

52 NJOY2016
3 RECONR LA-UR-17-20093

where Er is the resonance energy and ρ=ka depends on the channel radius or
the scattering radius as specified by NAPS. The phase shifts are given by

ϕ0 = ρ̂ , (19)
ϕ1 = ρ̂ − tan−1 ρ̂ , (20)
3ρ̂
ϕ2 = ρ̂ − tan−1 , (21)
3 − ρ̂2
15ρ̂ − ρ̂3
ϕ3 = ρ̂ − tan−1 , and (22)
15 − 6ρ̂2
105ρ̂ − 10ρ̂3
ϕ4 = ρ̂ − tan−1 , (23)
105 − 45ρ̂2 + ρ̂4

where ρ̂=kâ depends on the scattering radius. The final components of the cross
section are the actual line shape functions ψ and χ. At zero temperature,

1
ψ = , (24)
1 + x2
x
χ = , (25)
1 + x2
2(E − Er′ )
x = , (26)
Γr

and
Sℓ (|Er |) − Sℓ (E)
Er′ = Er + Γnr (|Er |) , (27)
2(Pℓ (|Er |)
in terms of the shift factors

S0 = 0 , (28)
1
S1 = − , (29)
1 + ρ2
18 + 3ρ2
S2 = − , (30)
9 + 3ρ2 + ρ4
675 + 90ρ2 + 6ρ4
S3 = − , and (31)
225 + 45ρ2 + 6ρ4 + ρ6
44100 + 4725ρ2 + 270ρ4 + 10ρ6
S4 = − . (32)
11025 + 1575ρ2 + 135ρ4 + 10ρ6 + ρ8

To go to higher temperatures, define

Γr
θ=r , (33)
4kT E
A

NJOY2016 53
LA-UR-17-20093 3 RECONR

where k is the Boltzmann constant and T is the absolute temperature. The line
shapes ψ and χ are now given by
√
π θx θ

ψ= θ ReW , , (34)
2 2 2

and √
π θx θ

χ= θ ImW , , (35)
2 2 2
in terms of the complex probability function (see quickw, wtab, and w, which
came from the MC2 code[37])

∞ 2
e−t
Z
2 i
W (x, y) = e−z erfc(−iz) = dt , (36)
π −∞ z−t

where z=x+iy. The ψχ method is not as accurate as kernel broadening (see

BROADR) because the backgrounds (which are sometimes quite complex) are
not broadened, and terms important for energies less than about 16kT /A are
neglected; however, the ψχ method is less expensive than BROADR. Previous
versions of RECONR included ψχ broadening for the SLBW and Adler-Adler
representations only. This version also allows the method to be used for MLBW
cases. The SLBW approach can produce negative elastic cross sections. If found,
they are set to a small positive value, and a count is accumulated for a diagnostic
in the listing file.

Multilevel Breit-Wigner Representation (MLBW) The Lubitz-Rose method

used for calculating Multi-Level Breit-Wigner cross sections (csmlbw) is formu-
lated as follows:
1
I+ l+s
π X X2 X
σn (E) = 2 ℓsJ
gJ |1 − Unn (E)|2 , (37)
k 1
ℓ s=|I− | J=|l−s|
2

with

X iΓnr
ℓJ
Unn (E) = e2iϕℓ − , (38)
r
Er′ − E − iΓr /2

where the other symbols are the same as those used above. Expanding the
complex operations gives

54 NJOY2016
3 RECONR LA-UR-17-20093

1
I+ l+s
π X X2
 X Γnr
X 2 2
σn (E) = gJ 1 − cos 2ϕℓ −
k2 1 r
Γr 1 + x2r
ℓ s=|I− | J=|l−s|
2
 X Γnr 2xr 2 
+ sin 2ϕℓ + , (39)
r
Γr 1 + x2r

where the sums over r are limited to resonances in spin sequence ℓ that have
the specified value of s and J. Unfortunately, the s dependence of Γ is not
known. The file contains only ΓJ =Γs1 J +Γs2 J . It is assumed that the ΓJ can be
used for one of the two values of s, and zero is used for the other. Of course,
it is important to include both channel-spin terms in the potential scattering.
Therefore, the equation is written in the following form:

 
π X X X Γnr 2 2
σn (E) = g J 1 − cos 2ϕ ℓ −
k2 r
Γr 1 + x2r
ℓ J
 X Γnr 2xr 2  
+ sin 2ϕℓ + + 2Dℓ (1 − cos 2ϕℓ ) , (40)
r
Γr 1 + x2r

where the summation over J now runs from

1 1
||I − ℓ| − | → I + ℓ + , (41)
2 2
and Dℓ gives the additional contribution to the statistical weight resulting from
duplicate J values not included in the new J sum; namely,

I+ 12 l+s I+ℓ+ 12
X X X
Dℓ = gJ − gJ (42)
s=|I− 12 | J=|l−s| J=||I−ℓ|− 12 |

I+ℓ+ 12
X
= (2ℓ + 1) − gJ . (43)
J=||I−ℓ|− 12 |

A case where this correction would appear is the ℓ=1 term for a spin-1 nuclide.
There will be 5 J values: 1/2, 3/2, and 5/2 for channel spin 3/2; and 1/2 and
3/2 for channel spin 1/2. All five contribute to the potential scattering, but the
file will only include resonances for the first three.
The fission and capture cross sections are the same as for the single-level

NJOY2016 55
LA-UR-17-20093 3 RECONR

option. The ψχ Doppler-broadening cannot be used with this formulation of the

MLBW representation.
However, there is an alternate representation available that does support ψχ
broadening:

σn = σp
XX nh Γnr Grℓ i
+ σmr cos 2ϕℓ − (1 − )+ ψ(θ, x)
r
Γr Γnr
ℓ
Hrℓ o
+(sin 2ϕℓ + ) χ(θ, x) , (44)
Γnr

where
1 X Γr + Γ r ′
Grℓ = Γnr Γnr′ , (45)
2 (Er − Er′ )2 + (Γr + Γr′ )2 /4
′
r ̸= r
Jr′ ̸= Jr
and
X Er − Er′
Hrℓ = Γnr Γnr′ . (46)
(Er − Er′ )2 + (Γr + Γr′ )2 /4
r′ ̸= r
Jr′ ̸= Jr
Nominally, this method is slower than the previous one because it contains a
double sum over resonances at each energy. However, it turns out that G and H
are slowly varying functions of energy, and the calculation can be accelerated by
computing them at just a subset of the energies and getting intermediate values
by interpolation. It is important to use a large number of r′ values on each side
of r. The GH MLBW method is implemented in csmlbw2.

Adler-Adler Representation (Adler-Adler) The multilevel Adler-Adler

representation (csaa) is defined for ℓ=0 only. It is useful for fissionable ma-
terials. The total cross sections are given by

4π
σt (E) = 2
sin2 ϕ0
k√
 
π E X 1
+ (Gr cos 2ϕ0 + Hr sin 2ϕ0 ) ψ(θ, x)
k2 r
νr

+ (Hr cos 2ϕ0 − Gr sin 2ϕ0 ) χ(θ, x)

A2 A3 A4
+ A1 + + 2 + 3 + B1 E + B2 E 2
, (47)
E E E

56 NJOY2016
3 RECONR LA-UR-17-20093

where

µr − E
x= , (48)
νr

and where νr is the resonance half-width (corresponds to Γ/2 in the Breit-Wigner

notation), µr is the resonance energy, Gr is the symmetric total parameter, Hr
is the asymmetric total parameter, and the Ai and Bi are coefficients of the total
background correction.
The fission and capture cross section both use the form

√ 
π E X 1
σx (E) = [Gr ψ(θ, x) + Hr χ(θ, x)]
k2 r
νr

A2 A3 A4
+ A1 + + 2 + 4 + B1 E + B2 E 2
, (49)
E E E

where the values of G, H, Ai , and Bi appropriate for the desired reaction are
used.
Doppler-broadening can be applied as for the SLBW case, except note that
Γr in Eq. 27 must be replaced with 2νr . Doppler-broadened Adler-Adler cross
sections are more accurate than SLBW cross sections because the background is
smoother. However, cross sections below about 16kT /A will still be inaccurate.
The Adler-Adler method is not used in modern evaluations.

Reich-Moore Representation (RM) The Reich-Moore representation as

implemented in subroutine csrmat is a multi-level formulation with two fission
channels; hence, it is useful for both structural and fissionable materials. The
cross sections are given by

2π X X n o
ℓJ
(50)


σt = g J 1 − Re U nm + 2d ℓJ 1 − cos(2ϕ ℓ ) ,
k2
ℓ J
π XX n i
ℓJ 2
(51)

σn = g J |1 − U nn | + 2d ℓJ 1 − cos(2ϕ ℓ ) ,
k2
ℓ J
4π X X X ℓJ 2
σf = gJ |Inc | , and (52)
k2 c
ℓ J
σγ = σt − σn − σf , (53)

where Inc is an element of the inverse of the complex R-matrix and

NJOY2016 57
LA-UR-17-20093 3 RECONR

h i
ℓJ
Unn = e2iϕℓ 2Inn − 1 . (54)

The elements of the R-matrix are given by

1/2 1/2
iX Γnr Γcr
ℓJ
Rnc = δnc − . (55)
2 r Er − E − 2i Γγr

In these equations, “c” stands for the fission channel, “r” indexes the resonances
belonging to spin sequence (ℓ, J), and the other symbols have the same mean-
ings as for SLBW or MLBW. Of course, when fission is not present, σf can
be ignored. The R-matrix reduces to an R-function, and the matrix inversion
normally required to get Inn reduces to a simple inversion of a complex number.
As in the MLBW case, the summation over J runs from

1 1
||I − ℓ| − | → I + ℓ + . (56)
2 2
The term dℓJ in the expressions for the total and elastic cross sections is used
to account for the possibility of an additional contribution to the potential scat-
tering cross section from the second channel spin. It is unity if there is a second
J value equal to J, and zero otherwise. This is just a slightly different approach
for making the correction discussed in connection with Eq. (43). Returning to
the I=1, ℓ=1 example given above, d will be one for J=1/2 and J=3/2, and it
will be zero for J=5/2.
ENDF-6 format RM evaluations can contain a parameter LAD that indicates
that these parameters can be used to compute an angular distribution for elastic
scattering if desired (an approximate angular distribution is still given in File 4
for these cases). The current version of RECONR has such a capability, and it
can be used with RM evaluations. Because of channel-spin issues, it works best
with RML evaluations. See below for a discusion of angular distributions.

Hybrid R-Function Representation (HRF) The Hybrid R-Function rep-

resentation treats elastic scattering as a multi-level cross section using formulas
similar to those given above for the Reich-Moore format in the case where fis-
sion is absent. The other reactions are treated with formulas similar to those of
the SLBW method. The main use for this format is to provide a better repre-
sentation of competitive reactions than is provided by any of the other formats
described above. This treatment can include a background R-function, tabulated
charged-particle penetrabilities, and optical model phase shifts. Following the

58 NJOY2016
3 RECONR LA-UR-17-20093

RM notation, the elastic cross section is given by

1
I+ l+s
π X X2 X
σn = 2 ℓsJ 2
gJ |1 − Unn | , (57)
k 1
ℓ s=|I− | J=|l−s|
2

where the U function is given by the scalar version of Eq. (54):

h 2 i
ℓsJ
Unn = e2iϕℓ ℓsJ
− 1 . (58)
Rnn

The R-function itself is given by

iX Γnr
ℓsJ
Rnn =1− 0
− i PℓsJ RℓsJ , (59)
2 r Er − E − 2i Γγr

where RℓsJ
0 is a (complex) background R function and PℓsJ is a penetrability
factor. The background R function can either be read in or set to zero. The
penetrability and shift factors are computed from the scattering radius or channel
radius as for SLBW. The phase shifts ϕℓsJ can be computed from the scattering
radius as before, or the (complex) phase shifts can be read in from an optical
model calculation.
Note that resonance parameters are given explicitly for all three quantum
numbers ℓ, s, and J. No correction to the potential scattering cross section from
repeated J values is needed.
Elastic angular distributions can also be computed from HRF parameters if
the LAD parameter is set; however, RECONR does not support that.

Reich-Moore-Limited Representaton (RML) The Reich-Moore-Limited

representation is a more general multilevel and multichannel formulation. In
addition to the normal elastic, fission, and capture reactions, it allows for inelas-
tic scattering and Coulomb reactions. Furthermore, it allows resonance angular
distributions to be calculated. It is also capable of computing derivatives of
cross sections with respect to resonance parameters. See ERRORR. The RML
processing in NJOY is based on the SAMMY code[36]. The calculation in RE-
CONR makes use of several subroutines exported by the samm module; namely,
s2sammy, ppsammy, rdsammy, cssammy, and desammy.
The quantities that are conserved during neutron scattering and reactions
are the total angular momentum J and its associated parity π, and the RML
format lumps all the channels with a given J π into a “spin group.” In each spin

NJOY2016 59
LA-UR-17-20093 3 RECONR

group, the reaction channels are defined by c = (α, ℓ, s, J), where α stands for
the particle pair (masses, charges, spins, parities, and Q-value), ℓ is the orbital
angular momentum with associated parity (−1)ℓ , and s is the channel spin (the
vector sum of the spins of the two particles of the pair). The ℓ and s values must
vector sum to J π for the spin group. The channels are divided into incident
channels and exit channels. Here, the important input channel is defined by the
particle pair neutron+target. There can be several such incident channels in
a given spin group. The exit channel particle pair defines the reaction taking
place. If the exit channel is the same as the incident channel, the reaction is
elastic scattering. There can be several exit channels that contribute to a given
reaction.
The R-matrix in the Reich-Moore “eliminated width” approximation for a
given spin group is given by
X γλc γλc′
Rcc′ = + Rcb δcc′ , (60)
Eλ − E − iΓλγ /2
λ

where c and c′ are incident and exit channel indexes, λ is the resonance index
for resonances in this spin group, Eλ is a resonance energy, γλc is a resonance
amplitude, and Γλγ is the “eliminated width,” which normally includes all of the
radiation width (capture). The channel indexes runs over the “particle channels”
only, which doesn’t include capture. The quantity Rcb is the “background R-
matrix.”
In order to calculate the contribution of this spin group to the cross sections,
we first compute the following quantity:

1/2
X
Xcc′ = Pc1/2 L−1
c Ycc−1
′′ Rc′′ c′ Pc′ , (61)
c′′

where
Ycc′′ = L−1
c δcc′′ − Rcc′′ , (62)

and
Lc = Sc − Bc + iPc . (63)

Here, the Pc and Sc are penetrability and shift factors, and the Bc are boundary
constants. The cross sections can now be written down in terms of the Xcc′ . For
elastic scattering

4π Xh 2 i r
X
2
i
σelastic = sin ϕ c (1 − 2Xcc ) − Xcc sin(2ϕ c ) + |Xcc′ | ,
kα2 π ′
J c

60 NJOY2016
3 RECONR LA-UR-17-20093

where Xcc
r is the real part of X ′ , X i is the imaginary part, ϕ is the phase
′ cc cc′ c
shift, the sum over J π is a sum over spin groups, the sum over c is limited to
incident channels in the spin group with particle pair α equal to neutron+target,
and the sum over c′ is limited to exit channels in the spin group with particle
pair α. Similarly, the capture cross section becomes

4π X X Xh X i
σcapture = gJα X i
cc − |Xcc ′|
2
, (64)
kα2 π c c ′
J c

where the sum over J π is a sum over spin groups, the sum over c is a sum over
incident channels in the spin group with particle pair α equal to neutron+target,
and the sum over c′ includes all channels in the spin group. The cross sections
for other reactions (if present) are given by

4π X Xh X i
σreaction = gJα X i
cc − |Xcc′ |2
, (65)
kα2 π c ′
J c

where the sum over c is limited to channels in the spin group J π with particle
pair α equal to neutron+target, and the sum over c′ is limited to channels in
the spin group with particle pair α′ . The reaction is defined by α → α′ . This
is one of the strengths of the RML representation. The reaction cross sections
can include multiple inelastic levels with full resonance behavior. They can
also include cross sections for outgoing charged particles, such as (n,α) cross
sections, with full resonance behavior. The total cross section can be computed
by summing up its parts.
For non-Coulomb channels, the penetrabilities P , shift factors S, and phase
shift ϕ are the same as those given for the SLBW representation, except if a Q
value is present, ρ must be modified as follows:
r
−3 A A+1
ρ = (2.196771×10 ) |E + Q| . (66)
A+1 A

These factors are a little more complicated for Coulomb channels. See the
SAMMY reference for more details.
The RML representation is new to the ENDF format, and it isn’t represented
by any cases in ENDF/B-VII.0. There are experimental evaluations for 19 F and
35 Cl from ORNL available. However, the RML approach provides a very faithful
representation of resonance physics, and it should see increasing use in the future.

NJOY2016 61
LA-UR-17-20093 3 RECONR

RML Angular Distributions. One of the physics advances available when

using the RML format is the calculation of angular distributions from the reso-
nance parameters. A Legendre representation is used:

dσαα′ X
= BLαα′ (E)PL (cos β) , (67)
dΩCM
L

where the subscript αα′ indicates the cross section as defined by the two particle
pairs, PL is the Legendre polynomial of order L, and β is the angle of the outgoing
particle with respect to the incoming neutron in the CM system. The coefficients
BLαα′ (E) are given by a complicated six level summation over the elements of
the scattering matrix U , where

Ucc′ = Ωc Wcc′ Ωc′ , (68)

where
Ωc = ei(wc −ϕc ) , (69)

and
W = I + 2iX , (70)

with I being the identity matrix and X having been already given in Eq. 61.
The coefficients B become

1 XXXXXX 1
BLαα′ (E) = 2
4kα (2i + 1)(2I + 1)
A B C D E F
× Gc1 c′1 ;c2 c′2 ;L Re[(δc1c′1 − Uc1 c′1 )(δc2 c′2 − Uc∗2 c′ )] . (71)
2

The spins I and i are for the target and projectile for particle pair α. The
complex expressions for the geometric coefficient G are given in the SAMMY
documentation. The six summations are as follows:

A sum over spin groups defined by J1π

B sum over spin groups defined by J2π
C sum over entrance channels c1 belonging to group J1π with particle pair α
D sum over exit channels c′1 belonging to group J1π with particle pair α′
E sum over entrance channels c2 belonging to group J2π with particle pair α
F sum over exit channels c′2 belonging to group J2π with particle pair α′

Fig. 3 shows the first few Legendre coefficients for the elastic scattering cross
sections as computed by NJOY from the experimental evaluation for 19 F.

62 NJOY2016
3 RECONR LA-UR-17-20093

0.5

0.4

0.3

Coefficient 0.2

0.1

-0.0

-0.1

-0.2

-0.3

-0.4

-0.5
104 105 106
Energy (eV)

Figure 3: Legendre coefficients of the angular distribution for elastic scattering in 19 F using
the RML resonance representation (P1 solid, P2 dashed, P3 dotted).

Although first introduced in NJOY2012, computing resonance angular dis-

tributions is a new, little used feature, and so it is not enabled by default. To
activate it, change Want_Angular_Dist to true. The Legendre coefficients are
written into a section of File 4 on the RECONR PENDF file. Because the nor-
mal ENDF File 4 sections are not copied to the PENDF File 4, the presence of
File 4 on a PENDF file can be detected by subsequent modules, such a ACER
or GROUPR, and the resonance angular distributions can be used to replace the
ENDF File 4 values over the resonance energy range. The default in NJOY2016
is to use the conventional RM processing path for RM parameters. However,
there is an option to convert the RM parameters into RML format and process
them with the RML methods. If this is done, resonance angular distributions
can be computed for an RM evaluation. Change Want_SAMRML_RM to true.

Infinitely-Dilute Unresolved Range Parameters Infinitely dilute cross

sections in the unresolved-energy range are computed in csunr1 or csunr2 using
average resonance parameters and probability distributions from File 2. With
the approximations used, these cross sections are not temperature dependent;
therefore, the results are a good match to resolved resonance data generated

NJOY2016 63
LA-UR-17-20093 3 RECONR

using tempr>0. The formulas used are based on the SLBW approximation with
interference.

2π 2 X gJ  2 2
(72)

σn (E) = σp + Γn R n − 2Γn sin ϕ ℓ ,
k2 D
ℓ,J
2π 2 X gJ
σx (E) = Γn Γx Rx , and (73)
k2 D
ℓ,J
4π X
σp = (2ℓ + 1) sin2 ϕℓ , (74)
k2
ℓ

where x stands for either fission or capture, Γi and D are the appropriate average
widths and spacing for the ℓ,J spin sequence, and Ri is the fluctuation integral
for the reaction and sequence (see gnrl). These integrals are simply the averages
taken over the chi-square distributions specified in the file; for example,
 
Γn Γf
Γn Γf Ri =
Γ
Z Z Z
= dxn Pµ (xn ) dxf Pν (xf ) dxc Pλ (xc ) (75)
Γn (xn ) Γf (xf )
× , (76)
Γn (xn ) + Γf (xf ) + Γγ + Γc (xc )

where Pµ (x) is the chi-square distribution for µ degrees of freedom. The integrals
are evaluated with the quadrature scheme developed by R. Hwang for the MC2 -2
code[38] giving

Qµi Qνj
Wiµ
X X X
Rf = Wjν Wkλ . (77)
i j k
Γn Qµi + Γf Qνj + Γγ + Γc Qλk

The Wiµ and Qµi are the appropriate quadrature weights and values for µ degrees
of freedom, and Γγ is assumed to be constant (many degrees of freedom). The
competitive width Γc is assumed to affect the fluctuations, but a corresponding
cross section is not computed. The entire competitive cross section is supposed
to be in the File 3 total cross section as a smooth background.
It should be noted that the reduced average neutron width Γ0n (AMUN) is
given in the file, and
√
Γn = Γ0n E Vℓ (E) , (78)

64 NJOY2016
3 RECONR LA-UR-17-20093

where the penetrabilities for the unresolved region are defined as

V0 = 1 , (79)
ρ2
V1 = , and (80)
1 + ρ2
ρ4
V2 = . (81)
ρ + 3ρ2 + ρ4

Other parameters are defined as for SLBW.

Unresolved resonance parameters can be given as independent of energy, with
only fission widths dependent on energy, or as fully energy dependent. The first
two options are processed in csunr1, and the last one is processed in csunr2.

3.5 Code Description

RECONR is implemented as a public subroutine reconr exported by the Fortran-
90 module reconm defined by reconr.f90.
The first step is to read cards 1, 2, and 3 of the user’s input. The TAPEID
record of the input file (nendf) is read and printed, then the new TAPEID record
is written to the output file (npend). RECONR is now ready to enter the loop
over the desired materials.
For each material, space is allocated for the energy nodes (enode), and ruin
is called to read cards 4 through 7 of the user’s input. If the reconstruction tem-
perature (tempr) is greater than zero, a table of ψ and χ functions is generated
(the W table is used; see wtab and quickw). The findf utility routine from
module endf is then used to find the first card of File 1 (MF=1, MT=451) for
the desired material.
File 1 on the input ENDF file is examined to obtain certain constants and
flags and to analyze the directory (anlyzd). Subroutine anlyzd determines which
reactions should be considered “redundant”; that is, the reactions that are sums
of other reactions and will be included on the output PENDF file. The total
cross section (MT=1 for neutrons, MT=501 for photons) will always be included;
the nonelastic cross section (MT=3) will be included if it is needed for photon
production (that is, MF=12, MT=3 is found); the inelastic cross section (MT=4)
will be included if sections with MT in the range 51 – 91 occur in the file, and
the total fission reaction (MT=18) will be called redundant if the partial fission
representation (MT=19, 20, 21, 38) is found. MT103 (n,p) can be a summation
reaction if its partials MT600, MT601, ..., are present, and the same for the other
charged-particle absorption reactions. Space for the new material directory is

NJOY2016 65
LA-UR-17-20093 3 RECONR

then allocated (mfs, mts, ncs). Section identification and card counts will be
entered into these arrays as they are determined.
File 2 on the ENDF file is now checked using s2sammy (which was imported
from the samm module) to see whether the sammy method is needed. This de-
pends on whether RML resonance parameters are found, and whether conversion
to RML format has been requested for Reich-Moore or Breit-Wigner data (see
Want_SAMRML_RM and Want_SAMRML_BW). The variable nmtres flags the use of
SAMMY processing. Because RML evaluations can include more than the nor-
mal elastic, fission, and capture reactions, a list of reactions identified is printed.
Here is an example from an experimental 19 F evaluation.

resonance range information

----------------------------------
ier energy-range lru lrf method
1 1.000E-05 1.000E+06 1 7 sammy

samm resonance reactions: 2 102 51 52

samm max legendre order: 7
generating File 4 for resonance angular distributions

The next step is to read File 2, which contains resolved and unresolved reso-
nance parameters (if any). The array res is allocated to contain the File 2 data
and rdfil2 is called to read them. This routine uses the additional routines
rdf2bw, rdf2aa, rdf2hy, rdsammy, rdf2u0, rdf2u1, and rdf2u2 to read the dif-
ferent types of resonance parameters. The subroutine rdsammy is imported from
the samm module. In addition, another imported routine ppsammy is used to
prepare for the SAMMY calculation. While the resonance parameters are being
stored, RECONR adds each resonance energy to its list of energy nodes (enode).
In the unresolved energy range, RECONR uses the energies of tabulated param-
eters or fission widths if available. If the evaluation uses energy-independent
parameters, or if the energy steps between the nodes are too large (see wide),
rdfil2 creates additional node energies at a density of approximately 13 points
per decade (see egridu). Note that regions where the unresolved representation
for an element overlaps the resolved or smooth ranges are found and marked by
negative energy values. The energy nodes are sorted into order and duplications
are removed.
If the SAMMY method is active, and if angular distributions have been re-

66 NJOY2016
3 RECONR LA-UR-17-20093

quested (see Want_Angular_Dist), the maximum Legendre order defined by the

resonance data is printed out.
If unresolved data are present, subroutine genunr is called to compute the
infinitely-dilute unresolved average cross sections on the unresolved energy grid
using csunr1 or csunr2. Any backgrounds on File 3 are included, except in re-
gions of resolved-unresolved or unresolved-smooth overlap. The computed cross
sections are arranged in the order required by the special section with MF=2
and MT=152, which is written onto the PENDF tape by recout. Using the
normal ENDF style, this format is defined by the following:

[MAT,2,152/ZA,AWR,LSSF,0,0,INTUNR]HEAD
[MAT,2,152/0.,0.,5,1,NW,NUNR/
E1,STOT1,SELAS1,SFIS1,SCAP1,STRN1,
E2,STOT2,SELAS2,SFIS2,SCAP2,STRN2,
...
ENUNR,STOTNUNR,....]LIST

where NW = 6 + 6*NUNR. The definitions of the energy and cross section entries
are fairly obvious, except STRN stands for the current-weighted total cross section.
This format is specialized to “infinite dilution.” The more general form used for
self-shielded effective cross sections will be described in the UNRESR, PURR,
and GROUPR chapters of this manual.
The subroutine lunion is used to linearize and unionize the ENDF data.
Space is reserved for two buffers to be used by loada/finda, for the linearization
stack (x and y), and for the ENDF scratch area (scr). The length of the stack
(ndim) determines the smallest possible subdivision of each panel (energy points
as close as 2−ndim times the panel width can be generated). Since the number
of energies in the union grid may soon exceed the capacity of any reasonable
memory array, the existing list of energy nodes is copied to binary scratch storage
(the loada/finda system). This storage system consists of the buffers bold
and bnew and the scratch units iold and inew. The energy grid points will
“ping-pong” back and forth between units 14 and 15 as the union grid is built
up. Subroutine lunion now starts with MT=2 and checks each reaction in
sequence to determine whether the current grid (on iold) is sufficient to represent
the reaction to within the desired tolerance using linear interpolation. If not,
RECONR adds additional points by adaptively halving the intervals. The new
grid is stored on inew. The units inew and iold are swapped, and the next

NJOY2016 67
LA-UR-17-20093 3 RECONR

MT is processed. When all nonredundant reactions have been examined, the list
of energies in loada/finda storage is the desired linearized and unionized grid.
The storage used is deallocated.
This grid is used as the starting point for resonance reconstruction in resxs.
Subroutine resxs first reserves space for the loada/finda buffers bufr and bufg,
the linearization stack (x and y), and the partial cross sections (sig). The
length of the stack (ndim) determines the smallest possible subdivision of a panel
between two nodes (energy points as close a 2−ndim times the panel width can
be generated). Subroutine resxs then examines the grid on ngrid (iold from
lunion) panel by panel. Grid points are added and cross sections computed
until the convergence criteria discussed in Section 3.3 are satisfied. The cross
sections are copied to nout using loada, and resxs continues to the next panel.
This procedure is continued until all panels are converged. The result is a tape
(nout) containing the energy grid in the resonance region and the total, elastic,
fission, capture, and possibly additional cross sections at each energy point.
Unionization is obtained automatically in the resonance region since all of the
partials are computed simultaneously in sigma, using csslbw for SLBW parame-
ters, csmlbw for MLBW parameters, csaa for multi-level Adler-Adler parameters,
csrmat for Reich-Moore parameters, cshyb for Hybrid R-Function parameters,
cssammy for Reich-Moore-Limited parameters, and sigunr for unresolved reso-
nance parameters. This last routine retrieves the cross sections from the table
prepared by genunr. Subroutine cssammy is imported from the samm module.
A special feature of RECONR is the ability to reconstruct the cross sections at
tempr by ψχ broadening if SLBW or Adler-Adler parameters are given. This can
also be done for MLBW using the GH method implemented by csmlbw2. The
Doppler-broadened resonance shapes are obtained using quickw (see description
in the UNRESR chapter), and the linearization procedure proceeds as before.
The resonance cross sections on ngrid are merged with the ENDF cross
sections in emerge. First, the background grid from lunion is merged with the
resonance grid from resxs and written onto the loada/finda file, which will
accumulate the total cross section and any other redundant reactions required
(iold/inew). A loop is then set up over all nonredundant reactions. For each
grid point, the ENDF background cross section is obtained by interpolation. If
this grid point has a resonance contribution on nres, it is added. The resulting
net cross section at this point is added into the appropriate redundant cross
sections on iold/inew and also saved on ngrid. When all the energies for this
reaction have been processed, the cross sections on ngrid are converted into a

68 NJOY2016
3 RECONR LA-UR-17-20093

TAB1 record and written on nscr. This loop is continued until all reactions have
been processed. When emerge is finished, nscr contains cross sections for all the
nonredundant reactions, and iold contains the redundant summation reactions.
Control now passes to recout, which writes the new File 1 comments and dic-
tionary. It also writes a default version of the section with MF=2 and MT=151
that gives no resonance parameters. The upper limit of the resolved energy
range, eresh, is added to the “C2” field of the third card so that BROADR
knows not to broaden into the unresolved energy range. For materials with
unresolved data, a specially formatted section (MF=2, MT=152) is written con-
taining the infinitely-dilute unresolved cross sections. This section can be used
by BROADR and GROUPR to correct for resolved-unresolved overlap effects,
if necessary. Subroutine recout then steps through the reactions on nscr and
iold. Redundant summation reactions are converted to TAB1 records and in-
serted in the correct order. Nonredundant reactions are simply copied. Finally,
a MEND record is added and control is returned to reconr.
Now reconr either directs that this process be repeated for another isotope or
writes a TEND record and terminates. The result is a new file in ENDF format
containing the desired pointwise cross sections. Normally, only Files 1, 2, 3, 10,
and 13 are included for neutron files. However, if angular distribution processing
has been requested, a File 4 containing the Legendre coefficients will also be
written. Because the original ENDF File 4 was not copied to the PENDF file,
the presence of sections of File 4 on the PENDF file provides a flag to subsequent
modules that resonance angular distributions have been calculated. Only Files
1 and 23 are included for a photon file.
The SAMMY method is implemented in a separate Fortran-90 module samm
defined by sammy.f90. It exports the subroutines cssammy (computes cross sec-
tions, angular distributions, and derivatives), s2sammy (scans File 2 to see if
SAMMY method is needed and measure some sizes), ppsammy (sets up SAMMY
calculation), rdsammy (reads in File 2 data with optional conversion of BW or
RM data to RML form), and desammy (cleans up after the SAMMY calcula-
tion). It also exports some logical parameters, namely, Want_Partial_Derivs,
Want_Angular_Dist, Want_SAMRML_RM and Want_SAMRML_BW. See ERRORR for
the use of derivatives. If conversion from BW and/or RM was requested, it is
possible to get the resulting File 2 values printed out for checking. Just set imf2
in sammy.f90 to 1.
The cssammy subroutine uses abpart to compute some energy-independent
pieces of the cross sections and derivatives. The main work for cross sections,

NJOY2016 69
LA-UR-17-20093 3 RECONR

angular distributions, and derivatives in done in cross. The results for cross
sections are returned in sigp to be consistent with the other “cs” routines in RE-
CONR. Angular distributions are returned in siga, and sensitivities are returned
in sigd (derivatives of cross sections with respect to parameters). Subroutine
cross starts by initializing the quantities being calculated (cross sections, maybe
angular distributions, maybe derivatives), and then it sets up a loop over the
spin-parity groups. It initializes the results for this spin group and then calls
setr to compute the elements of the R-matrix (see Eq. 60 and other quantities,
such as the Y matrix, penetrabilities (rootp), and phase shifts. It then inverts
the Y matrix and calculates the X matrix of Eq. 61. See setxqx. It can then use
the X matrix to compute the contributions to the cross sections (sectio), maybe
angular distributions (setleg), and maybe derivatives from this spin group and
add them into the sum over groups. When the loop over spin groups is complete,
it normalizes things properly and returns its results.
Going back to subroutine setr, it computes the R-matrix first. It then com-
putes the phase shift, penetrabilities, and shift factors. For non-Coulomb cases,
the phase shifts come from sinsix, and the penetrability P and boundry con-
dition (S − B + iP )−1 come from pgh. For Coulomb cases, subroutine pghcou
computes all of these quantities. The penetrability P is converted to rootp for
use in Eq. 61.

3.6 Input Instructions

The input instructions for each module are given in the code as comment cards
at the beginning of the source code for each module. The RECONR instructions
are reproduced here for the convenience of the reader.

!---input specifications (free format)---------------------------

!
! card 1
! nendf unit for endf tape
! npend unit for pendf tape
! card 2
! tlabel 66 character label for new pendf tape
! delimited with quotes, ended with /.
! card 3
! mat material to be reconstructed
! ncards number of cards of descriptive data for new mf1
! (default=0)

70 NJOY2016
3 RECONR LA-UR-17-20093

! ngrid number of user energy grid points to be added.

! (default=0)
! card 4
! err fractional reconstruction tolerance used when
! resonance-integral error criterion (see errint)
! is not satisfied.
! tempr reconstruction temperature (deg kelvin)
! (default=0)
! errmax fractional reconstruction tolerance used when
! resonance-integral error criterion is satisfied
! (errmax.ge.err, default=10*err)
! errint maximum resonance-integral error (in barns)
! per grid point (default=err/20000)
! (note: the max cross section difference for
! linearization, errlim, and for reconstruction,
! errmin, are also tied to errint. to get maximum
! accuracy, set errint to a very small number.
! for economical production, use the defaults.)
! card 5
! cards ncards of descriptive comments for mt451
! each card delimited with quotes, ended with /.
! card 6
! enode users energy grid points
!
! cards 3, 4, 5, 6 must be input for each material desired
! mat=0/ terminates execution of reconr.
!
!-------------------------------------------------------------------

A sample input for processing an isotope from ENDF/B-VII follows (the line
numbers are for reference only and are not part of the input). First, mount the
ENDF/B-VII file for 235 U on unit 20.

1. reconr
2. 20 21/
3. ’pendf tape for U-235 from ENDF/B-VII’/
4. 9228 2/
5. .001/
6. ’92-U-235 from ENDF/B-VII’/
7. ’processed with NJOY’/

NJOY2016 71
LA-UR-17-20093 3 RECONR

8. 0/

Card 2 tells RECONR that the input ENDF tape will be on unit 20, and that
the output PENDF tape will be on unit 21. Card 3 is a “TAPEID” label for the
output PENDF file. Card 4 gives the MAT number for U-235 and says that two
additional comment cards will be given. Card 5 sets the reconstruction tolerance
to 0.1% (.001 as a fraction) with all its other parameters defaulted. Cards 6 and
7 are the two comment cards to be inserted into the PENDF files MF1/MT451
section. Finally, the “0/” terminates the RECONR input. The capability to
loop over multiple isotopes in RECONR is rarely used for neutron files, but it is
useful for photon interaction processing (see GAMINR). The resulting PENDF
tape will contain the desired TAPEID card, followed by 235 U data, a MEND
card and a TEND card.

3.7 Error Messages

error in reconr***illegal nsub for reconr

RECONR only processes sublibraries that contain cross section data. Check
whether the right input ENDF input tape was mounted.
error in anlyzd***too many redundant reactions
Increase the global parameter nmtmax=10.
error in xxxxxx***storage in enode exceeded
Too many energy nodes including the user’s nodes and the energies from
MF=2. Increase the global parameter nodmax=800000. This message can
come from rdf2bw, rdf2aa, rdf2hy, rdf2u0, rdf2u1, or rdf2u2.
error in xxxxxx***res storage exceeded
Too much resonance data. This should not occur for a conforming ENDF-
format file, because maxres is computed from the MF=2 line count in
the MF=1/MT=451 index. This message can come from rdfil2, rdf2bw,
rdf2aa, rdf2hy, rdf2u0, rdf2u1, or rdf2u2.
error in xxxxxx***storage in eunr exceeded.
Increase the global parameter maxunr=500. This message can come from
rdfil2, rdf2u1, or rdf2u2.
error in rdfil2***illegal resonance mode.
A resonance mode has been requested that RECONR does not understand.
error in rdf2bw***energy-dep scattering radius ...
This option is only used in MLBW for current evaluations.

72 NJOY2016
3 RECONR LA-UR-17-20093

message from rdf2bw***calc... of angular distribution not...

This option is only partially available in RECONR. This message can come
from rdf2bw (for Reich-Moore cases) or from rdf2hy (Hybrid R function).
error in rdf2hy***hybrid competing reactions not yet added
This option is not yet available in RECONR.
error in lunion***ill behaved threshold
The routine is having trouble adjusting the threshold to agree with the Q
value. Check the points near the threshold for this evaluation.
error in lunion***exceeded stack
Increase length of linearization stack ndim (currently 50).
error in resxs***stack exceeded
Increase length of reconstruction stack ndim (currently 50).
error in sigma***general r-matrix not installed.
This option is not yet available in RECONR.
error in sigma***illegal option.
There is a problem with the ENDF tape.
error in csmlbw***not coded for temperature gt 0 deg k
The ψχ Doppler-broadening doesn’t work for normal MLBW. This message
shouldn’t occur, because temperatures greater than zero will cause csmlbw2
to be called.
error in csrmat***not coded for temperature gt 0 deg k
The ψχ Doppler-broadening doesn’t work for RM. Use tempr=0. only.
error in cshybr***doppler broad’g not provided for hybrid
The ψχ Doppler-broadening doesn’t work for hybrid parameters. Use tempr=0.
only.
error in csaa***bad li value
There is an error in the evaluation format.
message from emerge–-negative elastic cross sects found
Negative elastic cross sections can occur for SLBW evaluations.
error in recout***for mf –- mt –-
Indexing and pair count for this section do not make sense.
calculation of angular distribution not installed -
Message comes from several resonance types that do not support the calcu-
lation of angular distributions. Some of them can be used if Want_SAMRL_RM
or Want_SAMRML_BW are true.
message from s2sammy***multiple isotopes... -
Multiple isotopes for RM sections don’t work with the SAMMY method.
The code automatically reverts to normal RM processing.

NJOY2016 73
LA-UR-17-20093 3 RECONR

error in s2sammy***res storage exceeded. -

Storage limited to maxres. Shouldn’t occur.
error in s2sammy***energy-dep scattering length... -
This only works for MLBW parameters.
error in rearrange***nres fault -
Trouble while rearranging resonances into spin-group order.
errorr in findsp***quantum numbers in file 2 do not... -
Problems with the quantum numbers for the evaluation.
error in checkqn***error in quantum numbers -
Problems with the quantum numbers for the evaluation.
error in lmaxxx***lllmax limit to 51 -
Problems with the Clebsch-Gordan coefficients for the angular distributions.
error in clbsch***did not count correctly -
Problems with the Clebsch-Gordan coefficients for the angular distributions.
error in pspcou***llmax larger than 100 -
Problem computing the Coulomb phase shifts.
error in bigeta***I0 sum failed -
Problem for the Coulomb routine.
error in bigeta***K0 sum failed -
Problem for the Coulomb routine.
error in bigeta***L1 sum failed -
Problem for the Coulomb routine.
error in bigeta***K1 sum failed -
Problem for the Coulomb routine.
error in setleg***nppx too large -
Problem generating Legendre polynomials.

3.8 Input-Output Units

The following logical units are used:

10 nscr1 in reconr, nout in lunion, and nin in emerge. Contains copy

of nonredundant sections from original ENDF tape.
11 nscr2 in reconr; ngrid in lunion, resxs, and emerge. Contains
union grid for ENDF tape (not counting resonances).
12 nscr3 in reconr, nout in resxs, and nres in emerge. Contains
resonance grid and cross sections.

74 NJOY2016
3 RECONR LA-UR-17-20093

13 nscr4 in reconr is used for two separate purposes. In resxs it is

a binary scratch file nscr used for the unthinned resonance data.
In emerge and recout, it is nmerge and contains the nonredundant
reactions on the union grid.
14/15 iold/inew in lunion. Are used locally only to accumulate union grid
for ENDF cross sections. Destroy after use.
14/15 iold/inew in emerge. Are used locally only to accumulate summa-
tion cross sections on union grid.
20-99 User’s choice for ENDF (nendf) and PENDF (npend) tape numbers
to link RECONR with other NJOY modules.
5,6,7 See the NJOY chapter for a description of the I/O units.

Note that 11, 12, 14, and 15 are always binary. Unit 10 has the same mode as
nendf. Unit 13 is binary when used in RESXS, and it has the same mode as
npend elsewhere. npend can have a different mode than nendf.

3.9 Storage Allocation

Storage allocation in RECONR is sensitive to (1) the amount of resonance pa-
rameter data, (2) the number of energy-grid node values, (3) the size of the
resonance reconstruction stack, (4) the use of ψχ broadening, and (5) the sizes
of the loada/finda buffers. Other storage requirements are minor.
Buffer sizes can be reduced or increased at will. The result is a storage/speed
tradeoff with no change in capability or accuracy. See the global parameters
nbufg=2000, nbufr=2000, and nbuf=2000 at the beginning of the reconm mod-
ule.
The ψχ broadening option requires 7688 words of additional storage. There-
fore, memory use can be reduced if ψχ is not required. No code changes are
needed — just avoid tempr greater than zero.
Resonance reconstruction in resxs uses 5×ndim words. The parameter ndim
determines the smallest subdivision of a panel that can be obtained. Using
ndim=30 allows points to be generated with spacing as small as one-billionth of
the panel size (230 ).
The code currently allows for nodmax=800000 energy nodes and maxunr=500
unresolved points. These values tend to increase as additional very detailed
resonance evaluations appear, but the current values seem to be sufficient for
the evaluations existing as of this writing.

NJOY2016 75
LA-UR-17-20093 3 RECONR

76 NJOY2016
4 BROADR LA-UR-17-20093

4 BROADR
BROADR module generates Doppler-broadened cross sections in PENDF for-
mat starting from piecewise linear cross sections in PENDF format. The input
cross sections can be from RECONR or from a previous BROADR run. The
code is based on SIGMA1[27] by D. E. Cullen. The method is often called “ker-
nel broadening” because it uses a detailed integration of the integral equation
defining the effective cross section. It is a fully accurate method, treating all res-
onance and non-resonance cross sections including multilevel effects. BROADR
has the following features:

• An alternate calculation is used for low energies and high temperatures that
corrects a numerical problem of the original SIGMA1. (This problem has
been corrected in another way in later versions of SIGMA1.)
• Dynamic storage allocation is used, which allows the code to be run on
large or small machines with full use of whatever storage is made available.
• Reactions are broadened in parallel on a union grid, with the top of the
resolved resonance range being the typical upper limit for Doppler broad-
ening.
• The union grid is constructed adaptively to give a linearized representation
of the broadened cross section with tolerances consistent with those used in
RECONR. Energy points may be added to or removed from the input grid
as required for the best possible representation. Precision up to 9 significant
figures is allowed for energies.
• The summation cross sections such as total, nonelastic, and sometimes fis-
sion or (n,2n) are reconstructed to equal the sum of their parts.
• Standard thermal cross sections, integrals, and ratios are computed when
the temperature is 293.6K (0.0253 eV).
• The file directory (actually an index to the reactions present) is updated.
This chapter describes the BROADR module in NJOY2016.0.

4.1 Doppler-Broadening Theory

The effective cross section for a material at temperature T is defined to be that
cross section that gives the same reaction rate for stationary target nuclei as the
real cross section gives for moving nuclei. Therefore,
Z
ρvσ(v, T ) = dv′ ρ |v − v′ | σ(|v − v′ |) P (v′ , T ) , (82)

NJOY2016 77
LA-UR-17-20093 4 BROADR

where v is the velocity of the incident particles, v′ is the velocity of the target,
ρ is the density of target nuclei, σ is the cross section for stationary nuclei, and
P (v′ , T ) is the distribution of target velocities in the laboratory system. For
many cases of interest, the target motion is isotropic and the distribution of
velocities can be described by the Maxwell-Boltzmann function

α3/2 2
P (v′ , T ) dv′ = 3/2
exp(−αv ′ ) dv′ , (83)
π
where α = M/(2kT ), k is Boltzmann’s constant, and M is the target mass.
Eq. 82 can be partially integrated in terms of the relative speed V = |v − v′ |
to give the standard form of the Doppler-broadened cross section:

∞
α1/2
Z n o
2 2
σ(v) = 1/2 2 dV σ(V ) V 2 e−α(V −v) − e−α(V +v) . (84)
ϕ v 0

It is instructive to break this up into two parts:

σ(v) = σ ∗ (v) − σ ∗ (−v) , (85)

where
∞
α1/2
Z
∗ 2
σ (v) = 1/2 2 dV σ(V ) V 2 e−α(V −v) . (86)
π v 0

The exponential function in Eq. (86) limits the significant part of the integral to
the range
4 4
v−√ <V <v+√ .
α α
For σ ∗ (−v), the integral depends only on velocities satisfying

4
0≤V < √ .
α

These results can be converted to energy units using

1 4
2 16kT
Em = m √ = .
2 α A

Some examples are given in Table 1. Doppler-broadening effects will be impor-

tant below this energy and for any features such as resonances, thresholds, or
artificial discontinuities in evaluations that are not slowly varying with respect
√
to 2 Em E. As an example, for 235 U at 100 eV, Doppler effects are important
for features smaller than about 0.8 eV.

78 NJOY2016
4 BROADR LA-UR-17-20093

Table 1: Energy Parameter for Effective Doppler-Broadening

Target Temperature Energy Parameter (Em )

2H 300K 0.2 eV
235 U 300K 0.0017 eV
235 U 1.0 keV 69 eV

The numerical evaluation of Eq. (86) developed for SIGMA1 assumes that
the cross section can be represented by a piecewise linear function of energy
to acceptable accuracy. This is just the form of the NJOY PENDF files (see
√ √
RECONR). Defining the reduced variables y = αx and x = αV , the cross
section becomes
σ(x) = σi + si (x2 − x2i ) , (87)

with slope si = (σi+1 − σi )/(x2i+1 − x2i ). Eq. (86) can now be written as

N Z xi+1
∗ 1 X 2
X
σ(x) x2 e−(x−y) dx = Ai [σi − si x2i ] + Bi si , (88)

σ (y) =
π 1/2 y 2 i=0 xi i

where

x0 = 0 ,
xN +1 = ∞ ,
1 2
Ai = H2 + H1 + H0 , and
y2 y
1 4
Bi = H4 + H3 + 6H2 + 4yH1 + y 2 H0 ,
y2 y

and where Hn is shorthand for Hn (xi −y, xi+1 −y). The extrapolations to zero
and infinity assume a constant cross section (s0 =sN =0). The H functions are
the incomplete probability integrals defined by
Z b
1 2
Hn (a, b) = √ z n e−z dz . (89)
π a

These functions can be computed in two ways. First,

Hn (a, b) = Fn (a) − Fn (b) , (90)

NJOY2016 79
LA-UR-17-20093 4 BROADR

where Z ∞
1 2
Fn (a) = √ z n e−z dz . (91)
π a

These functions satisfy a recursion relation that can be used to obtain

1
F0 (a) = erfc(a) , (92)
2
1
F1 (a) = √ exp(−a2 ) , and (93)
2 π
n−1
Fn (a) = Fn−2 (a) + an−1 F1 (a) , (94)
2

where erfc(a) denotes the complementary error function

Z ∞
2 2
erfc(a) = √ e−z dz . (95)
π a

However, when Fn (a) ≈ Fn (b), the difference in Eq. (90) may lose significance.
In such cases, Hn (a, b) can be computed by a second method based on a direct
Taylor expansion of the defining integral. Write
Z b Z a
1 n −z 1 2
Hn (a, b) = √ z e dz − √ z n e−z dz = Gn (b) − Gn (a) . (96)
π 0 π 0

But by Taylor’s Theorem,

b−a ′ (b − a)m (m)

Gn (b) − Gn (a) = Gn (a) + ... + Gn (a) + ... . (97)
1! m!

Also,
dm−1  n −x2  2
G(m)
n (x) = x e = e−x Pnm (x) , (98)
dxm−1
where Pnm (x) is a polynomial with recursion relation

d m−1
Pnm (x) = P (x) − 2xPnm−1 (x) , (99)
dx n

with Pn1 = xn . From this point, it is straightforward to generate terms until the
desired number of significant figures is obtained.
When interpreting BROADR output, it is useful to remember several impor-
tant features of the Doppler-broadening process. A 1/v cross section remains
unchanged. Contrary to “popular knowledge”, the area under a resonance does
not remain unchanged unless E ≫ kT /A. In fact, each resonance develops a
new 1/v tail. Finally, a constant cross section (for example, elastic scattering)

80 NJOY2016
4 BROADR LA-UR-17-20093

Figure 4: The (n,α) cross section for 10 B from ENDF/B-V for three different temperatures
showing that a 1/v cross section is invariant under Doppler-broadening.

develops a 1/v tail at low energies after Doppler-broadening. These effects are
shown in Figs. 4, 5, and 6; they can be best understood by noting that the
Doppler process preserves reaction rate vσ(v) according to Eq. (82), and a finite
reaction rate is expected for T > 0K even as v → 0.
Very early (1980s) versions of BROADR and SIGMA1 assumed that the in-
put energy grid from RECONR could also be used to represent the Doppler-
broadened cross section before thinning. The grid was then thinned to take
advantage of the smoothing effect of Doppler broadening. Unfortunately, this
assumption is inadequate. The reconstruction process in RECONR places many
points near the center of a resonance to represent its sharp sides. After broad-
ening, the cross section in this energy region becomes rather smooth; the sharp
sides are moved out to energies where RECONR provides few points. At still
higher energies, the resonance line shape returns to its asymptotic value, and the
RECONR grid is adequate once more. The more recent versions of BROADR
check the cross section between points of the incoming energy grid, and add ad-
ditional grid points if they are necessary to represent the broadened line shape
to the desired accuracy. This effect is illustrated in Fig. 7.

NJOY2016 81
LA-UR-17-20093 4 BROADR

Figure 5: The elastic cross section for carbon from ENDF/B-V showing that Doppler-
broadening a constant cross section adds a 1/v tail.

Figure 6: The (n,γ) cross section for 240 Pu for several temperatures showing the effects of
Doppler broadening on resonances. The temperatures are 0K (solid), 30 000K
(dotted), and 300 000K (dash-dot). The higher resonances behave in the classical
manner even at 30 000K; note that the line shape returns to the asymptotic value
in the wings of the resonance. All resonances at 300 000K (and to a lesser extent
the first resonance for 30 000K) show the additional 1/v component that appears
when kT /A is large with respect to the resonance energy.

82 NJOY2016
4 BROADR LA-UR-17-20093

Figure 7: An expanded plot of the 20 eV resonance from Fig. 6 showing both thinning
and “thickening” of the energy grid produced adaptively by BROADR. The two
curves show the capture cross section at 0K and 300 000K. Note that the high-
temperature curve has fewer points than the 0K curve near the peak at 20 eV
and more points in the wings near 15 eV and 25 eV. Clearly, using the 0K grid
to represent the broadened cross section in the wings of this resonance would give
poor results.

4.2 Thermal Quantities

In thermal-reactor work, people make very effective use of a few standard thermal
constants to characterize nuclear systems. These parameters include the cross
sections at the standard thermal value of 0.0253 eV (2200 m/s), the integrals of
the cross sections against a Maxwellian distribution for 0.0253 eV, the g-factors
(which express the ratio between a Maxwellian integral and the corresponding
thermal cross section), η, α, and K1. Here, η is the Maxwellian-weighted average
¯ f )/(σf +σc ), α is the average of σc /σf , and K1 is the average of (ν̄ −1)σf −
of (ν)σ
σc . If BROADR is run for a temperature close to 293.6K (which is equivalent to
0.0253 eV), these thermal quantities are automatically calculated and displayed.
Here is a sample output for 235 U from ENDF/B-VII:

thermal quantities at 293.6 K = 0.0253 eV

-----------------------------------------
thermal fission xsec: 5.8490E+02
thermal fission nubar: 2.4367E+00
thermal capture xsec: 9.8665E+01

NJOY2016 83
LA-UR-17-20093 4 BROADR

thermal capture g-factor: 9.9086E-01

thermal capture integral: 8.6639E+01
capture resonance integral: 1.4043E+02
thermal fission integral: 5.0605E+02
thermal fission g-factor: 9.7628E-01
thermal alpha integral: 1.6828E-01
thermal eta integral: 2.0859E+00
thermal k1 integral: 6.4040E+02
equivalent k1: 7.2262E+02
fission resonance integral: 2.7596E+02
-----------------------------------------

4.3 Data-Paging Methodology

A piecewise linear representation of a reaction cross section of a resonance mate-
rial may require a very large number of energy points. For example, ENDF/B-
VII 238 U (MAT9237) requires 167 000 points for the total cross section for 0.1%
precision (errmax=err). It is impractical to load all these points into memory
simultaneously. However, the discussion following Eq. (86) in the theory section
shows that only a limited energy range around the point of interest is required.
The strategy used is to stage the cross-section data into three “pages” of npage
points each. Points in the center page can then be broadened using the npage or
√ √
more points on each side of the point of interest. If v − 4/ α and v + 4/ α are
both included in the three-page range, accurate broadening can be performed. If
not, a diagnostic warning is printed; the user should repeat the calculation with
a smaller temperature step or a larger page size.
There are many different reaction cross sections for each material. However,
the cross sections for high velocities are normally smooth with respect to 32kT /A
for any temperatures outside of stellar photospheres; therefore, they do not show
significant Doppler effects. Until recently the upper energy limit for Doppler
broadening was the smallest of (i) the input value thnmax, (ii) the upper limit
of the resolved-resonance energy range, (iii) the lowest threshold, or (iv) 1.0
MeV (the default input value for thnmax). No Doppler broadening or energy-
grid reconstruction is performed above that energy. In the past, and what users
typically expect as the default action, the second condition often set the Doppler
broadening upper limit.
However recent evaluated files have increasingly included threshold reactions
at energies within the resolved resonance energy range. For example, recent

84 NJOY2016
4 BROADR LA-UR-17-20093

JENDL evaluations for 235 U include a resolved-resonance range upper limit of

2.25 keV, but also include non-zero cross sections for an inelastic level with a
77 eV threshold. Under the rules itemized above, Doppler broadening of these
data stops at 77 eV. Other evaluations (e.g., ENDF/B-VI, ENDF/B-VII and
JEFF-3.1) share the same resolved-resonance range data but have zeroed out
this inelastic cross section from 77 eV to 2.25 keV and so Doppler broadening of
these files occurs throughout the resolved-resonance range, as most users expect.
Zeroing out non-zero data is a cludge from the past and so we have changed
the Doppler upper energy limit logic so that the top of the resolved resonance
range is now the default condition. This means that non-threshold reactions are
also Doppler broadened. The mathematics of this operation can produce non-
zero cross sections at energies below the reaction threshold. If this occurs those
cross sections are zeroed.
As noted in the BROADR module source code comments, users may specify
a negative value for thnmax to override these selection rules and force Doppler
broadening to an upper energy of abs(thnmax) eV. This has been a long-term
NJOY feature that remains unchanged in NJOY2016.
Finally, we note that the Ai and Bi factors in Eq. (88) depend only on the
energy (or velocity) values and not on the cross sections. Since the Ai and Bi
are expensive to compute, the code computes them only once for the points
of a unionized energy grid. The sum of Eq. (88) is accumulated for all the
non-threshold reactions simultaneously. This feature helps make BROADR run
faster.

4.4 Coding Details

The main subroutine for BROADR is broadr from module broadm. The code
begins by reading the user’s input (see Section 4.5). Storage is then allocated for
the loada/finda buffers (ibufo and ibufn) and for the scratch storage (iscr).
The buffer length nbuf can be changed at will (currently nbuf is 1000).
The input PENDF tape is searched for the desired material (mat1). If the
restart option is set (istart=1), the temperatures less than or equal to temp1
for mat1 are assumed to have been broadened previously, and they are copied to
the output file. In either case, the files for temp1 are copied to a scratch file on
unit nscr1.
Next, nscr1 is rewound and examined reaction by reaction. The energy grid
from the total cross section (MT1) is saved on scratch storage using loada. If
the input tape has not been through RECONR, the BROADR module will still

NJOY2016 85
LA-UR-17-20093 4 BROADR

run, but at possibly reduced accuracy. The next low-threshold reaction (that is,
the next reaction with a threshold less than emin, which is currently 1 eV) is
located on nscr1. The energy points are retrieved from scratch file iold (12 or
13) using finda, the cross sections for this reaction are computed on this grid,
and the results are stored on scratch file inew (13 or 12) using loada. The units
for iold and inew are then exchanged, and the entire process is repeated for the
next low-threshold reaction.
The final result of this process is a list of nreac low-threshold-reaction types
in mtr (usually MT2, MT18, and MT102), the threshold value for the first high-
threshold reaction (or the input value) in thnmax, and scratch file iold containing
the energy grid and all the low-threshold reactions (there are n2in points).
Now that the number of reactions to be broadened simultaneously is known
(nreac), storage for data paging can be assigned. The total amount of storage
available is namax. The value of namax should be as large as possible (current
value is 15 000 000). This space is divided up into the largest possible page
size, npage. An overflow region nstack is also allocated. Now that the page size
is known, the code allocates three pages for energies (e), three pages for each
reaction cross section (s), one extended page for the broadened energy grid (eb),
and three extended pages for the broadened cross section (sb). This system is
designed to use the available storage with maximum efficiency.
The cross sections on iold are now broadened by bfile3 (see below) and the
results are written on scratch unit inew using loada.
The directory from nscr1 is revised to reflect any thinning or thickening and
written on the output PENDF tape (nout). Note that the new temperature is
written into the first word of the Hollerith data record to simplify later searching.
The broadened cross sections are now converted into ENDF TAB1 records
and merged with the unbroadened cross sections on nscr1. The total cross
section (and sometimes nonelastic, inelastic, fission, (n,2n), or charged-particle
reactions) is reconstructed to equal the sum of its parts. The new Doppler-
broadened “MAT” on NOUT is a legal PENDF file with the same MAT number
as the original data but with a new temperature.
The process is now repeated for each of the ntemp2 final temperatures temp2
requested. Note that after each step inew contains the new data and iold
contains the previous data. If the “bootstrap” option is set (istrap=1), these
units are interchanged. For this option, stemp2(it) is always obtained from
temp2(it-1). Because of the thinning effect of Doppler-broadening, the broad-
ening runs faster at each step. The accumulation of error is usually not a prob-

86 NJOY2016
4 BROADR LA-UR-17-20093

lem. For istrap=0, temp1 is used for the starting temperature every time.
The broadening and energy-grid reconstruction are directed by bfile3 . The
routine loads data into the appropriate memory pages from scratch file iold and
then either calls broadn to broaden it (with thinning or thickening of the grid as
necessary) or calls thinb to thin it without broadening. The results are written
onto scratch file inew.
In broadn, the energy grid points just loaded into e by bfile3 are converted
to the dimensionless variables x and y [see Eq. (87)]. An adaptive reconstruction
of the Doppler broadened cross section is then performed for the energy range
in the center page using an inverted stack algorithm like the one described for
RECONR. The upper limit of each panel is taken to be a point from the input
grid, but in order to allow for thinning, up to nmax=10 of the input grid points
can be skipped before the actual upper limit is selected. In addition, the energy
of the upper limit cannot be more than step=2.01 times the energy of the
preceding point. The cross sections are now computed at the midpoint of the
top panel in the stack using bsigma. If the results differ from the values obtained
by interpolation by more than the specified tolerance, the new point is added to
the stack, and the tests are repeated. Otherwise, the top point in the stack is
converged. A backward check is made to see if some of the previous points can
be removed based on the new value, the new value is stored in the output array,
and the height of the stack is reduced by one. The routine now tries to subdivide
the new panel at the top of the stack in the same way. When the stack has been
reduced to one element, a new upper limit is chosen from the input energy grid
as described above, and the entire process is repeated. The reconstruction logic
in BROADR uses the same integral tests as RECONR. Refer to the RECONR
chapter for more details.
Subroutine bsigma is used to calculate the actual broadened cross section at
an energy point using the data in the three pages. First, the routine locates the
energy panel containing the desired energy en. It then loops over intervals below
the current point adding in contributions to σ̄ from the V −v term of Eq. (84)
until the contributions to the cross section become small. If the lower limit of
the bottom page is reached before convergence, a warning message is issued.
The routine then loops over intervals above the current point until convergence.
Once again, a warning is issued if necessary. Finally, the low-energy term [the
one involving V +v in Eq. (84)] is added, if applicable.
Subroutine thinb is provided for cases where the input cross section set is
to be thinned only. This routine uses the original SIGMA1 method. The first

NJOY2016 87
LA-UR-17-20093 4 BROADR

input point is always kept. The routine then loops over higher energy values.
For each grid point, all the points from there back to the last accepted point
are checked for their deviation from a straight line. If they all can be removed
without violating the specified tolerance, the interval is extended to the next
higher point and the tests are repeated. If any point in the range is too far from
the linear approximation, the last point in the range is accepted as an output
point, and the testing process is repeated starting from this new lower limit.
The procedure terminates when all of the points in the middle page have been
thinned, and control is returned to bfile3 to get the next page of data. Thinning
may have been a necessary feature in the past when computing resources were
limited but is a rarely used feature today.
Subroutine hunky has been modified from the original SIGMA1 version to
implement the alternate Hn (a, b) calculation when necessary (see hnabb). When
using the direct method, Fn values from the previous step are used in the differ-
ence of Eq. (90), and funky is called to get the new values. The Ai and Bi of
Eq. (88) are related to the s1 and s2 here.
Subroutine funky evaluates Fn (a) by the recursion formula of Eq. (94) using
the very accurate SLATEC version of the reduced complementary error function
from the NJOY2016 math module.
Function hnabb implements the alternate calculation described by Eqs. (96)-
(99). The series expansion is continued until about six significant figures are
guaranteed (see eps and hnabb). Currently, hnabb is called when only four
significant figures are reliable in hunky (see toler in hunky).

4.5 User Input

The following input instructions have been copied from the comment cards at
the start of BROADR.

!---input specifications (free format)---------------------------

!
! card 1
! nendf input endf tape (for thermal nubar only)
! nin input pendf tape
! nout output pendf tape
!
! card 2
! mat1 material to broadened and thinned
! ntemp2 number of final temperatures (default=1)

88 NJOY2016
4 BROADR LA-UR-17-20093

! istart restart (0 no, 1 yes, default 0)

! istrap bootstrap (0 no, 1 yes, default 0
! temp1 starting temperature from nin (default=0K)
!
! card 3
! errthn fractional tolerance for thinning
! thnmax max. energy for broadening and thinning
! (default=1 MeV)
! errmax fractional tolerance used when integral criterion
! is satisfied (same usage as in reconr)
! (errmax.ge.errthn, default=10*errthn)
! errint parameter to control integral thinning
! (usage as in reconr) (default=errthn/20000)
! set very small to turn off integral thinning.
! (A good choice for the convergence parameters
! errthn, errmax, and errint is the same set of
! values used in reconr)
!
! card 4
! temp2 final temperatures (deg Kelvin)
!
! card 5
! mat1 next MAT number to be processed with these
! parameters. Terminate with mat1=0.
!
!---input options------------------------------------------------
!
! The output tape will contain the ntemp2 final temperatures
! specified. It is necessary to have temp1.le.temp2(1).
! if temp2.eq.temp1, the data will be thinned only.
!
! restart Continue broadening an existing pendf tape. All
! temperatures are copied through temp1. Additional
! final temperatures are added by starting with the
! data at temp1.
!
! bootstrap If bootstrap is not requested, each final tempera-
! ture is generated by broadening directly from temp1
! to temp2. If bootstrap is requested, each final temp-
! erature is broadened from the preceding temperature.
! Bootstrapping is faster due to the thinning in the
! previous step. However, errors accumulate.

NJOY2016 89
LA-UR-17-20093 4 BROADR

!
! thnmax A possible upper limit for broadening and thinning.
! The actual upper limit is the lowest of (i) this input
! value; (ii) the end of the resolved resonance range;
! (iii) the lowest reaction threshold; or (iv) 1.0 MeV.
!
! A negative value for thnmax forces the Doppler
! broadening upper limit to be abs(thnmax) irrespective
! of the other conditions.
!
! Caution: this may cause one or more threshold
! reactions to be broadened. The magnitude of
! thnmax must be chosen to keep the number of
! broadenable reactions less than or equal to the
! maximum of ntt (160).
!
! Caution: for use in transport codes, it is recommended
! to use the program default. We don’t know how
! to compute the spectrum of scattered neutrons from
! a broadened inelastic level in the current generation
! of codes. Broadened cross sections for threshold
! reactions may be useful for other purposes.
!
!-------------------------------------------------------------------

Note that temp1 need not occur on nout if istart=0. The restart option
(istart=1) enables the user to add new temperatures to the end of an existing
PENDF tape. This option is also useful if a job runs out of time while processing,
for example, the fifth temperature in a job requesting six or more final temper-
atures. The job can be restarted from the nout. The first four temperatures
will be copied to the new nout and broadening will continue for temperature
five. The bootstrap option speeds up the code by using the broadened result for
temp2(i-1) as the starting point to obtain temp2(i). The thnmax parameters
can be used to speed up a calculation or to prevent the broadening of inappro-
priate data such as sharp steps or triangles in an evaluated cross section (for
example, ENDF/B-V lead).
The following example prepares a broadened PENDF file for 235 U from ENDF/B-
VII at two temperatures. The line numbers are for reference only; they are not
part of the input.

90 NJOY2016
4 BROADR LA-UR-17-20093

1. broadr
2. 20 21 22/
3. 9228 2/
4. .001/
5. 300. 1200./
6. 0/

On line 2, unit 20 should contain the ENDF file and unit 21 should a RECONR-
generated ASCII PENDF file of 0K cross sections for the isotope. Two materials
will be generated on unit 21 with 0.1% accuracy. First will be the 300K data,
followed by a MEND record, followed by the 1200K data, followed by MEND and
and TEND records. Best results are obtained when the error tolerance errthn
and the optional integral-thinning controls errmax and errint are the same as
those used for the RECONR run.

4.6 Error Messages

error in broadr***nin and nout must be same mode

Use coded to coded, or blocked binary to blocked binary. The latter is
faster due to the several tape copies performed in BROADR.
error in broadr***max. energy too large ...
The user requested Doppler broadening to an energy beyond the maximum
energy in the ENDF file.
error in broadr***too many low threshold reactions
The current limit is set by the global parameter ntt=180. Check tt, mtr,
and ntt in broadr, tt in bfile3, and sbt in broadn.
message from broadr–-desired mat and temp not on tape
Check the input PENDF file and the user input.
message from broadr–-no broadenable reactions
No low threshold reactions were found.
error in broadr***storage exceeded
Insufficient storage to update directory. Increase nwscr=1000 in broadr.
message from stounx–-sigma zero data removed ...
The input PENDF tape already contained a special unresolved section in
File 2. It has been removed. Rerun UNRESR if necessary.

NJOY2016 91
LA-UR-17-20093 4 BROADR

message from bsigma–-broadening truncated at a=––

The page is too small for the temperature difference requested. Increase
total storage available (namax) or repeat the calculation with smaller tem-
perature steps and istrap=1. The normal maximum size of a is 4.0 and a
is inversely proportional to Ti −Ti−1 .

4.7 Input/Output Units

The following units are used for input and output by BROADR.

10 nscr1 in BROADR. Contains the ENDF/B data at the initial tem-

perature.
12/13 iold/inew in BROADR. Contains union grid and low threshold re-
actions.
20-99 User’s choice for nin and nout to link with other modules.
Units 12 and 13 will always be binary. Unit 10 will have the same mode as nin
and nout.

4.8 Storage Allocation

All storage is divided in the most efficient way possible. The container array size
namax should be made as large as possible. The value of nbuf can be increased or
decreased at will — larger values will give faster execution. The value for nwscr
depends on the size of the ENDF/B dictionary, and 1000 words is sufficient for
all current evaluations.

92 NJOY2016
5 UNRESR LA-UR-17-20093

5 UNRESR
The UNRESR module is used to produce effective self-shielded cross sections for
resonance reactions in the unresolved energy range. In ENDF-format evalua-
tions, the unresolved range begins at an energy where it is difficult to measure
individual resonances and extends to an energy where the effects of fluctuations
in the resonance cross sections become unimportant for practical calculations. As
described in the ENDF format manual,[9] resonance information for this energy
range is given as average values for resonance widths and spacings together with
distribution functions for the widths and spacings. This representation can be
converted into effective cross sections suitable for codes that use the background
cross section method, often called the Bondarenko method,[39] using a method
originally developed for the MC2 code[37] and extended for the ETOX code[28].
This unresolved-resonance method has the following features:

• Flux-weighted cross sections are produced for the total, elastic, fission, and
capture cross sections, including competition with inelastic scattering.
• A current-weighted total cross section is produced for calculating the effec-
tive self-shielded transport cross section.
• The energy grid used is consistent with the grid used by RECONR.
• The computed effective cross sections are written on the PENDF tape in a
specially defined section (MF2, MT152) for use by other modules.
• The accurate quadrature scheme from the MC2-2 code[38] is used for com-
puting averages over the ENDF statistical distribution functions.

This chapter describes the UNRESR module in NJOY2016.0.

5.1 Theory
In the unresolved energy range, it is not possible to define precise values for the
cross sections of the resonance reactions σx (E), where x stands for the reaction
type (total, elastic, fission, or capture). It is only possible to define average
values. Of course, these average values should try to preserve the reaction rate:
Z E2
σx (E) ϕ0 (E) dE
∗ E1
σ 0x (E ) = Z E2
, (100)
ϕ0 (E) dE
E1

where ϕ0 (E) is the scalar flux, E ∗ is an effective energy in the range [E1 , E2 ],
and the range [E1 , E2 ] is large enough to hold many resonances but small with

NJOY2016 93
LA-UR-17-20093 5 UNRESR

respect to slowly varying functions of E. In order to calculate effective values

for the transport cross section, it is necessary to compute the current-weighted
total cross section also. It is given by
Z E2
σx (E) ϕ1 (E) dE
E1
σ 1t (E ∗ ) = Z E2
, (101)
ϕ1 (E) dE
E1

where the P1 component of the neutron flux, ϕ1 (E), is proportional to the neu-
tron current. To proceed farther, it is necessary to choose a model for the shape
of ϕℓ (E) in the vicinity of E ∗ . The model used in UNRESR is based on the B0
approximation for large homogeneous systems and narrow resonances:

C(E)
ϕℓ (E) = , (102)
[ Σt (E) ]ℓ
where C(E) is a slowly varying function of E, and Σt (E) is the macroscopic total
cross section for the system. In order to use this result in Eq. 100, it is further
assumed that the effects of other isotopes in the mixture can be approximated
by a constant called σ0 in the range [E1 , E2 ], or

C(E)
ϕℓ (E) = . (103)
[ σ0 + σt (E) ]ℓ
Therefore, the effective cross sections in the unresolved range are represented by
Z E2
σx (E)
C(E) dE
σ0 + σt (E)
σ 0x (E ∗ ) = ZEE1 2 , (104)
1
C(E) dE
E1 σ0 + σt (E)
with x being t for total, e for elastic, f for fission, and γ for capture, and
Z E2
σx (E)
C(E) dE
[ σ0 + σt (E) ]2
σ 1t (E ∗ ) = ZEE1 2 . (105)
1
C(E) dE
E1 [ σ0 + σt (E) ]2
This equation can also be written in the equivalent form

94 NJOY2016
5 UNRESR LA-UR-17-20093

Z E2
1
C(E) dE
∗ E1 σ0 + σt (E)
σ 1t (E ) = Z E2 − σ0 . (106)
1
2
C(E) dE
E1 [ σ0 + σt (E) ]

The parameter σ0 in Eq. 103 deserves more discussion. It can be looked at

as a parameter that controls the depth of resonance dips in the flux. When σ0 is
large with respect to the peak cross sections of resonances in σt (E), the shape of
the flux is essentially C(E). For smaller values of σ0 , dips will develop in the flux
that correspond to peaks in σt . These dips will cancel out part of the reaction
rate in the region of the peaks, thus leading to self-shielding of the cross section.
Analysis shows that it is possible to use this single parameter to represent the
effects of admixed materials or the effects of neutron escape from an absorbing
region. See the GROUPR chapter of this manual for additional details.
The cross sections that appear in the above integrals can be written as the
sum of a resonant part and a smooth part as follows:

XX
σx (E) = bx + σRx (E) = bx + σxsr (E−Esr ) , (107)
s r

where s is an index to a spin sequence, r is an index to a particular resonance in

that spin sequence, and Esr is the center energy for that resonance. The smooth
part bx can come from a smooth background given in the ENDF file, and it also
includes the potential scattering cross section σp for the elastic and total cross
sections (x=t and x=e). In terms of the smooth and resonant parts, the effective
cross sections become
Z E2
σRx (E)
C(E) dE
σ + σRt (E)
σ 0x (E ∗ ) = bx + ZEE1 2 , (108)
1
C(E) dE
E1 σ + σRt (E)
and
Z E2
1
C(E) dE
∗ E1 σ + σRt (E)
σ 1t (E ) = Z E2 − σ0 , (109)
1
2
C(E) dE
E1 [ σ + σRt (E) ]

where σ = bt + σ0 . It is convenient to transform the denominators of Eqs. 108

and 109 into

NJOY2016 95
LA-UR-17-20093 5 UNRESR

Z Z Z 
1 1 σt
C dE = C dE − C dE , (110)
σ + σt σ σ + σt
and

Z Z Z Z 
1 1 σt σσt
2
C dE = 2 C dE − C dE − C dE .
[σ + σt ] σ σ + σt [σ + σt ]2
(111)
Furthermore, since C(E) is assumed to be a slowly-varying function of E, it
can be pulled out through all integrals and dropped. The average cross sections
become

σI0x
σ 0x = bx + , (112)
1 − I0t
and

σI1t
σ 1t = bt + . (113)
1 − I0t − I1t
The last equation can also be written in the form
 
1 − I0t
σ 1t =σ − σ0 . (114)
1 − I0t − I1t
The average cross sections are thereby seen to depend on two types of “fluctuation
integrals:”
Z E2
1 σRx (E)
I0x = dE , (115)
E2 − E1 E1 σ + σRt (E)
and
Z E2
1 σσRt (E)
I1t = dE , (116)
E2 − E1 E1 [ σ + σRt (E) ]2
where x can take on the values t, n, f , or γ. Note that I1t ≤I0t , the difference
increasing as σ0 decreases from infinity.
Inserting the actual sums over resonances into the formula for I0x gives

E2 P
sr σxsr (E − Esr )
Z
1
I0x = P dE . (117)
E2 − E1 E1 σ + sr σtsr (E − Esr )
If the resonances were widely separated, only the “self” term would be important,
and one would obtain

96 NJOY2016
5 UNRESR LA-UR-17-20093

E2
σxsr (E − Esr )
Z
X 1
I0x = dE . (118)
sr
E2 − E1 E1 σ + σtsr (E − Esr )

Since the range of integration is large with respect to the width of any one
resonance, the variable of integration can be changed to ξ=E−Esr , and the
limits on ξ can be extended to infinity. For any one sequence, the interval
E2 −E1 is equal to the average spacing of resonances in that sequence times the
number of resonances in the interval. Therefore,

X 1 1 X Z ∞ σxsr (ξ)
I
I0x = dξ (119)
s
D s Ns r −∞ σ + σ tsr (ξ)

where Ds is the average spacing, and the “I” superscript indicates that this is the
“isolated resonance” result. Because there are assumed to be many resonances in
the interval, the sum over resonances can be changed to a multiple integration
over some characteristic set of parameters (such as widths) times the probability
of finding a resonance with some particular values of the parameters:

Z Z
1 X
fr = <f >s = dαPs (α) dβPs (β) · · · f (α, β, · · · ) . (120)
N r∈s

In the following text, this multiple integral (up to four fold) will be abbreviated
by writing the α integral only. The final results for isolated resonances are as
follows:

X 1 Z Z ∞
σxsα (ξ)
I
I0x = P (α) dξ dα , (121)
s
D s −∞ σ + σ tsα (ξ)

and

X 1 Z Z ∞
σσtsα (ξ)
I
I1t = P (α) 2
dξ dα . (122)
s
Ds −∞ [ σ + σtsα (ξ) ]

If the effects of overlap are too large to be neglected, overlap corrections to

the isolated resonance result can be constructed using the continued-fraction
generator
 
1 1 b
= 1− . (123)
a+b a a+b
Starting with the I0 integrals,

NJOY2016 97
LA-UR-17-20093 5 UNRESR

P
sr σxsr σxsr n
X
P = 1
σ + sr σtsr sr
σ + σtsr
X σtsr′ XX σts′ r′ o
− P − P . (124)
′
σ + σtsr ′ ′
σ + σtsr
r ̸=r s ̸=s r

Expand the second term in the braces to get

P
sr σxsr σxsr n
X
P = 1
σ + sr σtsr sr
σ + σtsr
X σtsr′
−
σ + σtsr + σtsr′
r′ ̸=r
n X σtsr′′ XX σts′ r′ o
1− P − P
′′
σ + σ tsr ′ ′
σ + σtsr
r ̸=r s ̸=s r
r ′′ ̸=r ′
XX σts′ r′ o
− P . (125)
σ + σtsr
s′ ̸=s r′

Neglecting the products of three different resonances in sequence s gives

P
sr σxsr σxsr
X
P =
σ + sr σtsr sr
σ + σtsr
n X σtsr′ o
× 1−
σ + σtsr + σtsr′
r′ ̸=r
h XX σts′ r′ i
× 1− P . (126)
′ ′
σ + σtsr
s ̸=s r

The factor before the opening brace is the isolated resonance result, the factor
in braces is the in-sequence overlap correction, and the factor in brackets is
the sequence-sequence overlap correction. Note that recursion can be used to
refine the sequence-sequence correction to any desired accuracy. Similarly, the
I1 integral requires

P
σσxsr
sr
X σσxsr h
= 1
[ σ + sr σtsr ]2 [ σ + σtsr ]2
P
sr
X σtsr′ XX σts′ r′ i2
− P − P . (127)
σ + σtsr σ + σtsr
r′ ̸=r ′ ′ s ̸=s r

98 NJOY2016
5 UNRESR LA-UR-17-20093

Once more, we expand the fraction and neglect terms that will result in products
of three or more different resonances in the same sequence. The result is

P
σσxsr
sr
X σσxsr
=
[ σ + sr σtsr ]2 2
P
sr [ σ + σtsr ]
X 2 o
n X σtsr′ σtsr′
× 1−2 +
′
σ + σtsr + σtsr′ σ + σtsr + σtsr′
r ̸=r r′ ̸=r
h XX σts′ r′ i
× 1− P , (128)
′ ′
σ + σtsr
s ̸=s r

where in-sequence and sequence-sequence overlap terms have been factored out.
The next step is to substitute these results back into the fluctuation integrals
I0 and I1 . The integrals over energy and the sums over different resonances in
each sequence can be handled as described above for isolated resonances. This
procedure will result in three different kinds of integrals. The first kind includes
the isolated resonance integrals already considered above

Z E2 X
1 σxsr
Bxs = dE
E2 − E1 E1 r σ + σtsr
Z Z ∞
1 σxsα (ξ)
= P (α) dξ dα , (129)
Ds −∞ σ + σtsα(ξ)

and

Z E2 X
1 σσtsr
Dts = dE
E2 − E1 E1 r [ σ + σtsr ]2
Z Z ∞
1 σσxsα (ξ)
= P (α) 2
dξ dα . (130)
Ds −∞ [ σ + σtsα (ξ) ]

Note that Dt ≤ Bt , the difference increasing as σ0 decreases from infinity.

The next kind are the in-sequence overlap integrals. The sum over r′ is
replaced by integrals over the probabilities of finding each partial width and the
probability of finding a resonance r′ at a distance η from resonance r.

NJOY2016 99
LA-UR-17-20093 5 UNRESR

Z E2 X X
1 σxsr σtsr′
V0xs = dE
E2 − E1 E1 r ′ σ + σtsr σ + σtsr + σtsr′
r ̸=r
Z Z Z Z
1 σxsα (ξ)
= P (α) P (β) Ω(η)
Ds2 σ + σtsα (ξ)
σtsβ (ξ − η)
dη dξ dβ dα , (131)
σ + σtsα (ξ) + σtsβ (ξ − η)

where ξ = E − Esr and η = Esr′ − Esr . Similarly,

Z E2 X X
1 σσtsr n σtsr′
V1ts = 2
2
E2 − E1 E1 r ′ [ σ + σtsr ] σ + σtsr + σtsr′
r ̸=r
 σtsr′ 2 o
− dE
σ + σtsr + σtsr′
Z Z Z Z
1 σσtsα (ξ)
= 2
P (α) P (β) Ω(η)
Ds [ σ + σtsα (ξ) ]2
n σtsβ (ξ − η)
2
σ + σtsα (ξ) + σtsβ (ξ − η)
h σtsβ (ξ − η) i2 o
− dη dξ dβ dα . (132)
σ + σtsα (ξ) + σtsβ (ξ − η)

The final class of integrals includes the sequence-sequence overlap corrections.

They are simplified by noting that the positions of resonances in different spin
sequences are uncorrelated. Therefore, Ω(η)=1, and the integral of the product
reduces to the product of the integrals.
Using the results and definitions from above, the fluctuation integrals become

X
I0x = Axs , (133)
s

h X i
Axs = (Bxs − V0xs ) 1 − Ats′ , (134)
s′ ̸=s

and

X h X i2
I1t = (Dts − V1ts ) 1 − Ats′ , (135)
s s′ ̸=s

where Eq. 134 provides a recursive definition of the Ats for the sequence-sequence
corrections as well as the normal value of Axs .

100 NJOY2016
5 UNRESR LA-UR-17-20093

These equations are formally exact for the sequence-sequence overlaps, but
in-sequence overlaps only include the interactions between pairs of resonances.
Three different approximations to this result are currently in use.

The MC2/ETOX Approximation The MC2 and ETOX codes use similar
approximations to the results above, except that MC2 does not include a calcu-
lation of the current-weighted total cross section. Both codes explicitly neglect
the in-sequence overlap corrections. This approximation was based on the as-
sumption that resonance repulsion would reduce the overlap between resonances
in a particular spin sequence, leaving the accidental close spacing of resonances
in different sequences as the dominant overlap effect. In addition, both codes
stop the recursion of Eq. 134 at At = Bt . Thus,

X  X 
I0x = Bxs 1 − Bts′ , (136)
s s′ ̸=s

and

X  X 2
I1t = Dts 1 − Bts′ . (137)
s s′ ̸=s

The equations for the effective cross sections in the MC2/ETOX approximation
become
X  X 
σ Bxs 1 − Bts′
s s′ ̸=s
σ 0x = bx + X  X  , (138)
1− Bts 1 − Bts′
s s′ ̸=s

and

X  X 2
σ Dts 1 − Bts′
s s′ ̸=s
σ 1t = bt + X  X  X X 2 , (139)
1− Bts 1 − Bts′ − Dts 1 − Bts′
s s′ ̸=s s s′ ̸=s

or

NJOY2016 101
LA-UR-17-20093 5 UNRESR

 X  X  
 1− Bts 1 − Bts′ 
s s′ ̸=s
 − σ0 . (140)
 
σ 1t = σ

X  X  X  X 2 
1 − Bts 1 − Bts −
′ Dts 1 − Bts 
′

s s′ ̸=s s s′ ̸=s

These are the equations that are used in the UNRESR module of NJOY. Note
that the equation in the ETOX code and report corresponding to Eq. 140 is
incorrect. The following equation was used in the ETOX code:
 
X X 
1−Bts 1 − Bts′
s s′ ̸=s
 
σ 1t = σ  − σ0 , (141)
  
X X
1 − C 1− C ′ 
ts ts
s s′ ̸=s

with Cts = Bts + Dts .

The MC2-2 Approximation The MC2-2 code includes the in-sequence over-
lap corrections, which the authors found to be more important than previously
thought. It uses additional approximations to obtain the equivalent of

X Bxs − V0xs
σ 0x = bx + σ . (142)
s
1 − Bts + V0ts

The additional approximations used are

1. Set Ats = Bts − V0ts (first-order sequence-sequence overlap),

2. Neglect the factor (1 − s′ ̸=s Ats′ ) in the denominator, and
P

3. Use the approximation 1 − i fi ≈ i (1 − fi ) on the numerator and de-

P Q

nominator.

These simplifications result in a loss of accuracy for the sequence-sequence

overlap correction at relatively low values of σ0 . The σ 1t term is not calculated.

The UXSR Approximation The experimental UXSR module was devel-

oped at Oak Ridge (with some contributions from LANL) based on coding from
the Argonne National Laboratory (ANL) in an attempt to include the sophis-
ticated in-sequence overlap corrections from MC2-2 without approximating the
sequence-sequence corrections so badly. It also implemented a calculation of the
current-weighted total cross section, which was omitted in MC2-2. The addi-
tional cost of using the full expressions for Eqs. 133 and 135 is minimal, and

102 NJOY2016
5 UNRESR LA-UR-17-20093

effective cross sections can be computed for lower values of σ0 when in-sequence
overlap is small (e.g., 238 U).

Now that expressions have been chosen for computing the cross sections in
terms of the isolated-resonance integrals, it is necessary to select an efficient
numerical method for computing them. The resonant parts of the cross sections
are given by
 
Γx
σxsr (E−Esr ) = σm ψ(θ, X) , (143)
Γ sr

and

σtsr (E−Esr ) = [ σm {cos 2ϕℓ ψ(θ, X) + sin 2ϕℓ χ(θ, X)} ]sr , (144)

where x takes on the values γ, f , or c for capture, fission, or competition, and

4πgJ Γn
σm = , (145)
k2 Γ
r
A
θ= Γ, (146)
4kT E0

2(E − E0 )
X= , (147)
Γ

2J + 1
gJ = , and (148)
2(2I + 1)

A √
k = 2.196771 × 10−3 E. (149)
1+A
The functions ψ and χ are the symmetric and antisymmetric components of the
broadened resonance line shape:
√  
θ π θX θ
ψ(θ, X) = ReW , , (150)
2 2 2
and

√
 
θX θ
χ(θ, X) = θ πImW , , (151)
2 2
where

W (x, y) = exp[−(x + iy)2 ] erfc[−i(x + iy)] (152)

NJOY2016 103
LA-UR-17-20093 5 UNRESR

is the complex probability integral. The methods for computing ψ and χ are
well known (see quikw).
The first integral needed is

Z Z
1 σxsα (ξ)
Bxs = P (α) dξ dα
Ds σ + σtsα (ξ)
Z Z
1 σm (Γx /Γ)ψ(θ, X)
= P (α) dξ dα
Ds σ + σm {cos 2ϕℓ ψ(θ, X) + sin 2ϕℓ χ(θ, X)}
Z Z
1 Γx ψ(θ, X)
= P (α) dX dα (, 153)
Ds 2 cos 2ϕℓ β + ψ(θ, X) + tan 2ϕℓ χ(θ, X)

where

σ
β= . (154)
σm cos 2ϕℓ
The second integral needed is

Z Z
1 σtsα (ξ)
Bts = P (α) dξ dα
Ds σ + σtsα ξ)
Z Z
1 Γ ψ(θ, X) + tan 2ϕℓ χ(θ, X)
= P (α) dX dα . (155)
Ds 2 β + ψ(θ, X) + tan 2ϕℓ χ(θ, X)

Both of these integrals can be expressed in terms of the basic J integral:

Z
1 Γ
Bxs = P (α) J(β, θ, tan 2ϕℓ , 0) dα , and
Ds cos 2ϕℓ
Z
1
Bts = P (α) Γ J(β, θ, tan 2ϕℓ , tan 2ϕℓ ) dα , (156)
Ds

where
Z ∞
1 ψ(θ, X) + b χ(θ, X)
J(β, θ, a, b) = dX . (157)
2 −∞ β + ψ(θ, X) + a χ(θ, X)
The D integral can be handled in the same way, but only total reaction is re-
quired.

104 NJOY2016
5 UNRESR LA-UR-17-20093

Z Z
1 σσtsα (ξ)
Dts = P (α) dξ dα
Ds [ σ + σtsα (ξ) ]2
Z Z
1 Γ βψ(θ, X) + tan 2ϕℓ χ(θ, X)
= P (α) dX dα
Ds 2 [ β + ψ(θ, X) + tan 2ϕℓ χ(θ, X) ]2
Z
1
= P (α) Γ K(β, θ, tan 2ϕℓ , tan 2ϕℓ ) dα , (158)
Ds

where
Z ∞
1 β [ ψ(θ, X) + b χ(θ, X) ]
K(β, θ, a, b) = dX . (159)
2 −∞ [ β + ψ(θ, X) + a χ(θ, X) ]2
A method for computing J, including the interference effects, has been de-
veloped by Hwang for MC2-2[38]. However, this method was not available in
the days when MC2 and ETOX were developed. Therefore, UNRESR uses only
J(β, θ, 0, 0) and K(β, θ, 0, 0) in computing the isolated-resonance fluctuation in-
tegrals. A direct integration is used over most of the X range, but the part of
the integral arising from large X is handled using analytic integrations of the
asymptotic forms of the arguments (see ajku).
The final step is to do the n-fold integration over the probability distributions
for the resonance widths. This integration has been abbreviated as a single inte-
gration over α in the above equations. The method used was originally developed
for MC2-2 and is based on Gauss-Jacobi quadratures. A set of 10 quadrature
points and weights is provided for each of the χ2 probability distributions with
1 through 4 degrees of freedom. These quadratures convert the n-fold integral
into an n-fold summation. The value of n can be as large as 4 when Γn , Γf , Γγ ,
and Γc (competitive width) are all present.
Although UNRESR neglects the effects of overlap between resonances in the
same spin sequence and the effects of interference in the elastic and total cross
sections, it still gives reasonable results for the background cross section values
needed for most practical problems. Modern evaluations are steadily reducing
the need for accurate unresolved calculations by extending the resolved resonance
range to higher and higher energies. Ultimately, UNRESR should be upgraded to
use the UXSR approach. Another alternative is to generate self-shielded effective
cross sections from ladders of resonances chosen statistically (see PURR). This
avoids many of the approximtions in the overlap corrections.
In NJOY2016, running the PURR module after UNRESR overwrites the
UNRESR output with the PURR results. In fact, UNRESR can be omitted

NJOY2016 105
LA-UR-17-20093 5 UNRESR

from the processing stream. To use UNRESR results, either omit PURR from
the processing stream or run it before running UNRESR.

5.2 Implementation
In implementing this theory in UNRESR, there are special considerations in-
volving the choice of an incident energy grid, what to do if the unresolved range
overlaps the resolved range or the range of smooth cross sections, the choice of
the σ0 grid, how to interpolate on σ0 , and how to communicate the results to
other modules.

Choice of Energy Grid. The same logic is used to choose the incident energy
grid in UNRESR and RECONR. It is complicated, because of the several different
representations available for unresolved data, and because of the existence of
evaluations that have been carried over from previous versions of ENDF/B or
ENDF/B-VII evaluations with inadequate energy grids. Even many modern
evaluations have inadequate energy grids.
For evaluations that give energy-independent unresolved-resonance param-
eters, there is still an energy dependence to the cross sections. Because this
dependence is normally somewhere between constant and a 1/v law, a fairly
coarse grid with about 13 points per decade should be sufficient to allow the
cross sections to be computed reliably using linear-linear interpolation.
If the evaluation uses energy-dependent parameters, the normal rule would
be to use the energies that were provided by the evaluator and to obtain inter-
mediate cross sections by interpolation. Unfortunately, some of the evaluations
carried over from earlier days contain some energy intervals that are quite large
(for example, steps by factors of 10). The evaluators for these isotopes assumed
that the user would use parameter interpolation and compute the cross sections
at a number of intermediate energies in these long steps. Even some newer
evaluations contain large jumps in the energy grid. UNRESR will detect such
evaluations and add additional energy points in the large energy steps using an
algorithm similar to the one used for the cases with energy-independent param-
eters. For NJOY2016, large jumps in the energy grid are any with step ratios
greater than wide, which is currently set to 1.26.
The final energy grid can be observed by scanning the printed output from
UNRESR.

106 NJOY2016
5 UNRESR LA-UR-17-20093

Resolved-Unresolved Overlap. Elemental evaluations include separate en-

ergy ranges in MF2/MT151 for each of the isotopes of the element, and these
energy ranges do not have to be the same for each isotope. This means that the
lower end of an unresolved range may overlap the resolved range from another
isotope, and the upper end of the unresolved range for an isotope can overlap
the smooth range of another isotope. These overlap regions are detected by UN-
RESR as the resonance data are read in, and they are marked by making the
sign of the incident energy value negative.

Choosing a σ0 Grid. There are two factors to consider, namely, choosing

values that will represent the shape adequately, and limiting the range of σ0 to
the region where the theory is valid. The σx (σ0 ) curves start out decreasing
from the infinite dilution value as 1/σ0 as σ0 decreases from infinity (1 × 1010
in the code). The curve eventually goes through an inflection point at some
characteristic value of σ0 , becomes concave upward, and approaches a limiting
value at small σ0 that is smaller than the infinite-dilution value. Decade steps
are often used, but the user should try to select values that include the inflection
point and not waste values on the 1/σ0 region. Half-decade values are useful
near the inflection point (e.g., 100, 300, 1000 for 235 U). The grid interval chosen
should be consistent with the interpolation method used (see below).
Choosing the lower limit for σ0 is a more difficult problem. As shown in
the theory section (5.1), resonance overlap effects are developed as a series in
1/σ0 , and the series is truncated after only one step of recursion in Eq. 134.
This means that the overlap correction should be most accurate for large σ0
and gradually get worse as σ0 decreases. Experience shows that the correction
can actually get large enough to produce negative cross sections for small σ0 .
(This problem can also show up as a failure in interpolation when a log scheme
has been selected.) For isotopes that have relatively narrow resonances spaced
relatively widely, such as 238 U, UNRESR gives reasonable results to σ0 values
as low as 0.1. For materials with stronger overlap, such as 235 U, a lower limit
around 100 is more reasonable. A few of the heavy actinide evaluations have
been seen to break down for σ0 values lower than 1000. This problem is not too
serious in practice. The fertile materials, which appear in large concentrations
in reactors, allow the necessary small values of σ0 . The fissile materials have to
be more dilute, and the larger σ0 limit needed for them is not usually a problem.
The UXSR approximation discussed above allows one to reach somewhat
smaller σ0 values.

NJOY2016 107
LA-UR-17-20093 5 UNRESR

Interpolating on σ0 . It turns out that these functions are difficult to interpo-

late because they have a limited radius of convergence. Although approximate
schemes have been developed based on using functions of similar shape such as
the tanh function[40], better results can be obtained by using different interpo-
lation schemes for the low- and high-σ0 ranges. The TRANSX-CTR code[12]
used interpolation in 1/σ0 for high σ0 , Lagrangian interpolation of lnσx vs lnσ0 ,
for intermediate values, and a σ02 extrapolation for very low σ0 . Unfortunately,
schemes like this sometimes give ridiculous interpolation excursions when the
polynomials are not suitable. Therefore, TRANSX-2[41] has had to revert to
using simple linear interpolation, which is always bounded and predictable, but
which requires relatively fine σ0 grids.

Communicating Results to Other Modules. NJOY tries to use ENDF-like

files for all communications between the different calculational modules. Because
the unresolved effective cross sections were originally derived from the resonance
parameters in File 2, it seemed reasonable to create a new section in File 2
to carry the unresolved cross sections onto other modules, and a special MT
number of 152 was selected for this purpose. RECONR creates an MT152 that
contains only the infinitely-dilute unresolved cross sections. UNRESR overwrites
it with self-shielded unresolved cross sections. GROUPR can then use these
cross sections in its calculation of the multigroup constants. The format used
for MT152 is given below using the conventional ENDF style.

[MAT,2,152/ ZA, AWR, LSSF, 0, 0, INTUNR ] HEAD

[MAT,2,152/ TEMZ, 0, NREAC, NSIGZ, NW, NUNR/
SIGZ(1), SIGZ(2),...,SIGZ(NSIGZ),
EUNR(1),
SIGU(1,1,1), SIGU(1,2,1),...,SIGU(1,NSIGZ,1),
SIGU(2,1,1),...
...
SIGU(NREAC,1,1),...,SIGU(NREAC,NSIGZ,1),
EUNR(2),...
<continue for energies through EUNR(NUNR)>
...SIGU(NREAC,NSIGZ,NUNR) ] LIST

where NREAC is always 5 (for the total, elastic, fission, capture, and current-
weighted total reactions, in that order), NSIGZ is the number of σ0 values, NUNR
is the number of unresolved energy grid points, and NW is given by

108 NJOY2016
5 UNRESR LA-UR-17-20093

NW=NSIGZ+NUNR*(1+5*NSIGZ)

5.3 User Input

The following summary of the user input instructions was copied from the com-
ment cards at the beginning of the UNRESR module in the NJOY2016 source
file.

!---input specifications (free format)---------------------------

!
! card 1
! nendf unit for endf tape
! nin unit for input pendf tape
! nout unit for output pendf tape
! card 2
! matd material to be processed
! ntemp no. of temperatures (default=1)
! nsigz no. of sigma zeroes (default=1)
! iprint print option (0=min, 1=max) (default=0)
! card 3
! temp temperatures in Kelvin (including zero)
! card 4
! sigz sigma zero values (including infinity)
! cards 2, 3, 4 must be input for each material desired
! matd=0/ terminates execution of unresr.
!
!--------------------------------------------------------------------

Card 1, as usual, specifies the input and output units for the module. The
input PENDF file on nin must have been processed through RECONR and
BROADR. It will contain default versions of the special unresolved section with
MF=2 and MT=152 that gives the infinitely-dilute unresolved cross sections for
each temperature. The output PENDF file nout will contain revised sections
MF2, MT152 that give effective cross sections vs σ0 for all temperatures.
Card 2 is used to specify the material desired (matd), the number of temper-
atures and background cross sections desired (ntemp and nsigz), and the print

NJOY2016 109
LA-UR-17-20093 5 UNRESR

option (iprint). The actual temperature and σ0 values are given on Cards 3 and
4. Temperatures should be chosen to be adequate for the planned applications.
The temperature dependence of the effective cross sections is usually monotonic
and fairly smooth. Polynomial interpolation schemes using T are often used,
and roughly uniform spacing for the temperature grid (or spacing that increases
modestly as T increases) is usually suitable.
The choice of a grid for σ0 is more difficult. The curves of cross section versus
σ0 have an inflection point, and it is important to choose the grid to represent
the inflection point fairly well. A log spacing for σ0 is recommended. Very small
values of σ0 should not be used. These considerations are discussed in more detail
in the “Implementation” section (5.2) above. Unfortunately, the best choice for
a grid can only be found by trial and error.

5.4 Output Example

The following portion of UNRESR output is for 238 U from ENDF/B-VII.0.

unresr...calculation of unresolved resonance cross sections 494.4s

storage 8/ 20000

unit for input endf/b tape ........... -21

unit for input pendf tape ............ -23
unit for output pendf tape ........... -24

temperatures ......................... 2.936E+02

4.000E+02
...
sigma zero values .................... 1.000E+10
1.000E+03
1.000E+02
5.000E+01
2.000E+01
1.000E+01
5.000E+00
2.000E+00
1.000E+00
5.000E-01
print option (0 min., 1 max.) ........ 1

110 NJOY2016
5 UNRESR LA-UR-17-20093

mat = 9237 temp = 2.936E+02 494.4s

energy = 2.0000E+04
1.433E+01 1.428E+01 ... 1.311E+01 1.297E+01 1.291E+01 1.288E+01
1.380E+01 1.375E+01 ... 1.264E+01 1.252E+01 1.247E+01 1.244E+01
0.000E+00 0.000E+00 ... 0.000E+00 0.000E+00 0.000E+00 0.000E+00
5.294E-01 5.280E-01 ... 4.654E-01 4.540E-01 4.491E-01 4.464E-01
1.433E+01 1.424E+01 ... 1.246E+01 1.230E+01 1.223E+01 1.220E+01
energy = 2.3000E+04
1.414E+01 1.411E+01 ... 1.307E+01 1.293E+01 1.288E+01 1.285E+01
1.364E+01 1.361E+01 ... 1.262E+01 1.250E+01 1.245E+01 1.242E+01
0.000E+00 0.000E+00 ... 0.000E+00 0.000E+00 0.000E+00 0.000E+00
4.962E-01 4.951E-01 ... 4.426E-01 4.325E-01 4.281E-01 4.257E-01
1.414E+01 1.407E+01 ... 1.246E+01 1.229E+01 1.223E+01 1.219E+01
...
energy = 1.4903E+05
1.140E+01 1.140E+01 ... 1.124E+01 1.118E+01 1.115E+01 1.113E+01
1.126E+01 1.126E+01 ... 1.110E+01 1.104E+01 1.101E+01 1.099E+01
0.000E+00 0.000E+00 ... 0.000E+00 0.000E+00 0.000E+00 0.000E+00
1.427E-01 1.427E-01 ... 1.406E-01 1.386E-01 1.374E-01 1.367E-01
1.213E+01 1.212E+01 ... 1.174E+01 1.156E+01 1.147E+01 1.141E+01
generated cross sections at 18 points 494.9s

The display of the cross section table for each energy has σ0 reading across (in
decreasing order) and reaction type reading down (in the order of total, elastic,
fission, capture, and current-weighted total). Four σ0 values were removed from
the table to make it narrower for this report.

5.5 Coding Details

The main entry point for UNRESR is subroutine unresr, which is exported by
module unresm. The subroutine starts by reading in the user’s input and output
unit numbers and opening the files that will be needed during the UNRESR
run. After printing the introductory timer line, storage is allocated for an array
scr, which will be used for reading in ENDF records. The default size of this
array is maxscr=1000, which has proved sufficient for all evaluations tried so far.
UNRESR now prints out the user’s unit numbers on the output listing and calls
uwtab to prepare internal tables used by uw to compute the real and imaginary
parts of the complex probability integral.
The next step is to read in the tapeid records of the ENDF and PENDF

NJOY2016 111
LA-UR-17-20093 5 UNRESR

tapes. The loop over materials starts at statement number 110 by reading in
the user’s values for the ENDF MAT number, number of temperatures ntemp,
number of σ0 values nsigz, and print flag iprint. If this is not the end of the
material loop (mat=0), the actual values of the temperatures and background
cross sections are read into the arrays temp and sigz. The input quantities are
echoed to the output listing to help detect input errors.
The code then begins a loop over the requested temperatures. It writes the
current values of material ID, temperature, and time on the output listing. It
then reads the resonance parameters from the section with MT=151 in File 2 of
the ENDF tape using rdunf2. The arrays eunr with length maxeunr=150 and
arry with length with length maxarry=10000 are used to store these data. Next,
it reads the background cross sections from File 3 for each of the resonance reac-
tions using rdunf3. Here, sb is used to store the data. The array b is allocated
with sufficient length to build the output record to be written in MT=152 on
the new PENDF file.
The next loop is over all the energy grid points at this temperature. The
grid points were determined in rdunf2, and the nunr points are stored in the
array eunr. The background cross sections are stored in an array sb(ie,ix).
The energy index takes on nunr values, and the reaction index ix takes on
four values. The calculation of the actual effective cross sections takes place in
unresl. The results for each energy appear in the array sigu(ix,is), where
is takes on nsigz values. Each unresl array is stored into the accumulating
output array b and printed on the output listing.
At this point, UNRESR checks to see if there is a previous version of MT152
on the PENDF tape. If so, these new data will replace it. If not, a new section
with MT=152 will be created. In either case, a new section MT451 in File 1
is generated containing the current temperature and the correct entry in the
directory for the PENDF tape. Finally, the new MT152 for this temperature is
copied onto the output file from the b array, and the rest of the contents of this
temperature on the old PENDF file are copied to the new PENDF file. After
writing a report on the number of resonance points generated and the amount of
computer time used, UNRESR branches back to continue the temperature loop.
When the last temperature has been processed, the code closes the files, writes
a final report on the output listing, and terminates.
Note that UNRESR takes special branches for materials with no resonance
parameters or materials with no unresolved parameters. Therefore, the user can
freely request an UNRESR run even when there are no unresolved resonance

112 NJOY2016
5 UNRESR LA-UR-17-20093

data present on the ENDF tape. UNRESR simply copies nin to nout in this
case.
The subroutine rdunf2 is used to read in the unresolved resonance parameters
from File 2 of the input ENDF tape. It is very similar to the corresponding coding
in RECONR. The array scr is used to read in the ENDF record, the resonance
parameter data are stored in the array arry, and the final grid of energy values
is stored in eunr. Note that rdunf2 will add extra energy nodes for evaluations
with energy-independent parameters or for energy-dependent evaluations that
have energy steps larger than the factor wide, which is currently set to 1.26.
The subroutine also discovers resolved-unresolved or unresolved-smooth overlap,
flags those energy values, and prints messages to the user on the output listing.
Subroutine ilist is used to insert a new energy into a list of energies in
ascending order. It is used to manage the accumulation of the grid of energy
nodes.
Subroutine rdunf3 is used to read in the background cross sections in the
unresolved range from File 3. The unresolved energy grid determined by rdunf2
is used for the background cross sections.
The main calculation of the effective cross sections for the unresolved range
is performed in subroutine unresl. The calculation is done in two passes: first,
the potential scattering cross section is computed; second, the unresolved cross
sections are computed. The passes are controlled by the parameter ispot. In
both cases, resonance parameter data stored in arry by rdunf2 are used. The
coding is similar to that used in RECONR down to the point where the ETOX
statistical averages are computed. The three loops over kf, kn, and kl carry
out the n-fold quadrature represented as integrals over α in the text. They
account for variations in the fission width, neutron width, and competitive width.
The capture width is assumed to be constant. Note that ajku is called in the
innermost loop to compute the J and K integrals in xj and xk, respectively.
The K integral returned by this routine is actually J + K in our notation. The
final quantities are computed in the loops over itp and is0. Note that tk is
equivalent to Bts +Dts in our notation. Similarly, tl is equivalent to Bxs , and tj
is equivalent to Bts . The last step in this subroutine is to compute the average
cross sections by summing over spin sequences. The loop over ks computes xj
as s Bts and xk as s Cts . With these quantities available, it is easy to finish
P P

the calculation of the effective cross sections as given by Eq. 138 and Eq. 140.
The subroutines uunfac, intrf, and intr are similar to corresponding rou-
tines from RECONR and don’t require further discussion here. Subroutine ajku

NJOY2016 113
LA-UR-17-20093 5 UNRESR

is used to compute the J and K integrals without interference corrections. The

subroutines quikw, uwtab, and uw implement a package for computing the com-
plex probability integral efficiently. It was originally developed at ANL for the
MC2 code. When it is used, a pair of 62 × 62 tables for the real and imagi-
nary parts of the complex probability integral are precomputed using uwtab and
uw. Values of W for small arguments are obtained by interpolating in these pre-
computed tables. Values of W for larger arguments are obtained using various
asymptotic formulas.

5.6 Error Messages

error in unresr***mode conversion between nin and nout

Input and output files must both be in ASCII mode (positive unit numbers),
or they must both be in binary mode (negative unit numbers).
error in unresr***storage exceeded
There is not enough room in the b array. Increase nb, which currently is
5000.
error in rdunf2***energy dependent data undefined
When using unresolved resonance formalisms with energy dependent pa-
rameters (e.g. the fission width, etc.), these data need to be defined over
the entire unresolved resonance region. In rare cases, such as ENDF/B-
VII.0 Pu239, this is not the case, leading to NaN cross section values. This
is an evaluation error.
error in rdunf2***storage exceeded
There is not enough room in arry. Increase the value of maxarry, which
currently is 10000.
error in unresl***storage exceeded
The code is currently limited to 50 spin sequences. For more spin sequences,
it will be necessary to increase the dimensions of several arrays in this
subroutine.

114 NJOY2016
6 HEATR LA-UR-17-20093

6 HEATR
The HEATR module generates pointwise heat production cross sections and
radiation damage energy production for specified reactions and adds them to
an existing PENDF file. The heating and damage numbers can then be easily
group averaged, plotted, or reformatted for other purposes. An option of use to
evaluators checks ENDF/B files for neutron/photon energy-balance consistency.
The advantages of HEATR include

• Heating and damage are computed in a consistent way.

• All ENDF/B neutron and photon data are used.
• ENDF/B-6 charged-particle distributions are used when available.
• Kinematic checks are available to improve future evaluations.
• Both energy-balance and kinematic KERMA factors can be produced.

This chapter describes the HEATR module in NJOY2016.0.

6.1 Theory of Nuclear Heating

Heating is an important parameter of any nuclear system. It may represent
the product being sold – as in a power reactor – or it may affect the design of
peripheral systems such as shields and structural components.
Nuclear heating can be conveniently divided into neutron heating and photon
heating (see Fig. 8). Neutron heating at a given location is proportional to the
local neutron flux and arises from the kinetic energy of the charged products
of a neutron induced reaction (including both charged secondary particles and
the recoil nucleus itself). Similarly, photon heating is proportional to the flux of
secondary photons transported from the site of previous neutron reactions. It is
also traceable to the kinetic energy of charged particles (for example, electron-
positron pairs and recoil induced by photoelectric capture).
Heating, therefore, is often described by KERMA[30] (Kinetic Energy Release
in Materials) coefficients kij (E) defined such that the heating rate in a mixture
is given by

XX
H(E) = ρi kij (E)ϕ(E) , (160)
i j

where ρi is the number density of material i, kij (E) is the KERMA coefficient
for material i and reaction j at incident energy E, and ϕ(E) is the neutron or

NJOY2016 115
LA-UR-17-20093 6 HEATR

NEUTRON GAMMA
FLUX FLUX

A Z
prompt
gammas

production delayed
and burnup gammas
A'
prompt and
delayed
prompt local delayed local non-local
heating heating heating

Figure 8: Components of nuclear heating. HEATR treats the prompt local neutron heating
only. Gamma heating is computed in GAMINR. Delayed local heating and photon
production are not treated by NJOY, and they must be added at a later stage.

photon scalar flux at E. KERMA is used just like a microscopic reaction cross
section except that its units are energy × cross section (eV-barns for HEATR).
When multiplied by a flux and number density, the result would give heating in
eV/s.
The “direct method” for computing the KERMA coefficient is

X
kij (E) = E ijℓ (E)σij (E) , (161)
ℓ

where the sum is carried out over all charged products of the reaction including
the recoil nucleus, and E ijℓ is the total kinetic energy carried away by the ℓth
species of secondary particle. These kinds of data are now becoming available for
some materials with the advent of ENDF/B-VI and later, but earlier ENDF/B
versions did not include the detailed spectral information needed to evaluate
Eq. 161.
For this reason, NJOY computes KERMA factors for many materials by the
“energy-balance method”[42]. The energy allocated to neutrons and photons is
simply subtracted from the available energy to obtain the energy carried away
by charged particles:

116 NJOY2016
6 HEATR LA-UR-17-20093

 
kij (E) = E + Qij − E ijn − E ijγ σij (E) , (162)

where Qij is the mass-difference Q-value for material i and reaction j, E n is the
total energy of secondary neutrons including multiplicity, and E γ is the energy
of secondary photons including photon yields.
This method was well suited for use with ENDF/B-V, or any other evalua-
tion containing neutron and photon spectral data, but not the particle spectra
required by the direct method. The disadvantage of this method is that the
KERMA factor sometimes depends on a difference between large numbers. In
order to obtain accurate results, care must be taken by the evaluator to ensure
that photon and neutron yields and average energies are consistent. In fact,
the lack of consistency in ENDF/B-V often revealed itself as negative KERMA
coefficients[43].
However, a negative KERMA coefficient is not always the defect it seems to
be. It must be remembered that heating has both neutron and photon com-
ponents. A negative KERMA might indicate that too much energy has been
included with the photon production in the evaluation. This will result in ex-
cessive photon heating if most of the photons stay in the system. However, the
negative KERMA will have just the right magnitude to cancel this excess heat-
ing. The energy-balance method guarantees conservation of total energy in large
homogeneous systems.
In this context, large and homogeneous means that most neutrons and pho-
tons stay in their source regions. It is clear that energy-balance errors in the
evaluation affect the spatial distribution of heat and not the total system heat-
ing when the energy-balance method is employed.
A final problem with the energy-balance method occurs for the elemental
evaluations common in earlier versions of ENDF/B. Isotopic Q-values and cross
sections are not available in the files. It will usually be possible to define adequate
cross sections, yields, and spectra for the element. However, it is clear that the
available energy should be computed with an effective Q given by
X
ρi σi Qi
i
Q= X , (163)
ρi σi
i

where ρi is the atomic fraction of isotope i in the element. This number is energy

NJOY2016 117
LA-UR-17-20093 6 HEATR

dependent and can be represented only approximately by the single constant

Q allowed in ENDF/B. HEATR allows the user to input an auxiliary energy-
dependent Q for elements.
For elastic and discrete-level inelastic scattering, the neutron KERMA coef-
ficient can be evaluated directly without reference to photon data. For other
reactions, conservation of momentum and energy can be used to estimate the
KERMA or to compute minimum and maximum limits for the heating. HEATR
includes an option that tests the energy-balance KERMA factors against these
kinematic limits, thereby providing a valuable test of the neutron-photon consis-
tency of the evaluation. If the energy-balance heating numbers for a particular
isotope should fail these tests, and if the isotope is important for a “small” sys-
tem, an improved evaluation is probably required. The alternative of making ad
hoc fixes to improve the local heat production is dangerous because the faults
in the neutron and/or photon data revealed by the tests may lead to significant
errors in neutron transport and/or photon dose and nonlocal energy deposition.
In practice, an exception to this conclusion must be made for the radiative
capture reaction (n,γ). The difference between the available energy E+Q and
the total energy of the emitted photons is such a small fraction of E+Q that it
is difficult to hold enough precision to get reasonable recoil energies. Moreover,
the emitted photons cause a component of recoil whose effect is not normally in-
cluded in evaluated capture spectra. Finally, the “element problem” cited above
is especially troublesome for capture because the available energy may change
by several MeV between energies dominated by resonances in different isotopes
of the element, giving rise to many negative or absurdly large heating num-
bers. These problems are more important for damage calculations (see below)
where the entire effect comes from recoil and the compensation provided by later
deposition of the photon energy is absent.
For these reasons, HEATR estimates the recoil due to radiative capture using
conservation of momentum. The recoil is the vector sum of the “kick” caused
by the incident neutron and the kicks due to the emission of all subsequent
photons. Assuming that all photon emission is isotropic and that the directions
of photon emission are uncorrelated, the photon component of recoil depends on
the average of Eγ2 over the entire photon spectrum

E Eγ2
ER = + , (164)
A + 1 2(A + 1)mc2

118 NJOY2016
6 HEATR LA-UR-17-20093

where mc2 is the neutron mass energy. The second term is important below
25 – 100 keV. This formula gives an estimate that works for both isotopes and
elements and has no precision problems. However, it does not explicitly conserve
energy, and isotopes with bad capture photon data can still cause problems.

6.2 Theory of Damage Energy

Damage to materials caused by neutron irradiation is an important design consid-
eration in fission reactors and is expected to be an even more important problem
in fusion power systems. There are many radiation effects that may cause dam-
age, for example, direct heating, gas production (e.g., helium embrittlement),
and the production of lattice defects.
A large cluster of lattice defects can be produced by the primary recoil nucleus
of a nuclear reaction as it slows down in a lattice. It has been shown that
there is an empirical correlation between the number of displaced atoms (DPA,
displacements per atom) and various properties of metals, such as elasticity.
The number of displaced atoms depends on the total available energy Ea and
the energy required to displace an atom from its lattice position Ed . Since the
available energy is used up by producing pairs,

Ea
DPA = . (165)
2Ed

The values of Ed used in practice are chosen to represent the empirical correla-
tions, and a wide range of values is found in the literature[44, 45, 46]. Table 2
gives the default values used in NJOY2016. The energy available to cause dis-
placements is what HEATR calculates. It depends on the recoil spectrum and
the partition of recoil energy between electronic excitations and atomic motion.
The partition function used is given by Robinson[47] based on the electronic
screening theory of Lindhard[48] (see Fig. 9).
The damage output from HEATR is the damage energy production cross
section (eV-barns). As in Eq. 160, multiplying by the density and flux gives
eV/s. Dividing by 2Ed gives displacements/s. This result is often reduced by an
efficiency factor (say 80%) to improve the fit to the empirical correlations.

NJOY2016 119
LA-UR-17-20093 6 HEATR

Table 2: Typical Values for the Atomic Displacement Energy Needed to Compute DPA[46].

Element Ed , eV Element Ed , eV
Be 31 Co 40
C 31 Ni 40
Mg 25 Cu 40
Al 27 Zr 40
Si 25 Nb 40
Ca 40 Mo 60
Ti 40 Ag 60
V 40 Ta 90
Cr 40 W 90
Mn 40 Au 30
Fe 40 Pb 25

Figure 9: Examples of the portion of the primary recoil energy that is available to cause
lattice displacements in metallic lattices. The remaining energy leads to electronic
excitation. The quantity plotted is P (E) from Eq. 200 divided by E. The 25 eV
cutoff is also discussed in connection with Eq. 200.

6.3 Computation of KERMA Factors By Energy Balance

6.3.1 The general case

The older ENDF/B files do not usually give photon production data for all
partial reactions. Summation reactions such as nonelastic (MT=3) and inelastic

120 NJOY2016
6 HEATR LA-UR-17-20093

(MT=4) are often used. It is still possible to compute partial KERMA factors
for those summation reactions by reordering Eq. 162 as follows:

X X
kij = n
kij (E) − E iℓγ σiℓ (E) , (166)
j∈J ℓ∈J

where j runs over all neutron partials contained in J, and ℓ runs over all photon
partials in J. The total KERMA is well defined, but partial KERMAS should
be used only with caution.
HEATR loops through all the neutron reactions on the ENDF/B tape. If
energy balance is to be used, it computes the neutron contributions needed for
the first term. These are
h i
n
kij (E) = E + Qij − E ijn (E) σij (E) . (167)

The Q-value is zero for elastic and inelastic scattering. For (n,n′ ) particle
reactions represented by scattering with an LR flag set, Q is the ENDF “C1”
field from MF=3. For most other reactions, Q is the “C2” field from MF=3.
HEATR allows users to override any Q-value with their own numbers.
The E n value as used in Eq. 167 is defined to include multiplicity. The
multiplicity is either implicit — for example, 2 for (n,2n) — or is retrieved from
the ENDF/B file (e.g. for the mt5 reaction). The average energy per neutron is
computed differently for discrete two-body reactions and continuum reactions.
For elastic and discrete inelastic scattering (MT=2, 51-90),

E  
En = 2
1 + 2Rf1 + R , (168)
(A + 1)2
where f1 is the center-of-mass (CM) average scattering cosine from MF=4 and
R is the effective mass ratio. For elastic scattering, R=A, but for threshold
scattering,
r
(A + 1)S
R=A 1− , (169)
AE
where S is the negative of the C2 field from MF=3.
For continuum scattering, the average energy per neutron is computed from
the secondary neutron spectrum, g, in MF=5 using

NJOY2016 121
LA-UR-17-20093 6 HEATR

Z U
E n (E) = E ′ g(E, E ′ ) dE ′ , (170)
0

where U is defined in MF=5. If g is tabulated (LAW=1 or LAW=5), the in-

tegral is carried out analytically for each panel by making use of the ENDF/B
interpolation laws. For the simple analytic representations (LAW=7, 9, or 11),
the average energies are known[9].
The neutron cross sections required by Eq. 167 are obtained from an existing
PENDF file (see RECONR, and BROADR).
When the neutron sum in Eq. 166 is complete, the code processes the photon
production files. If the evaluation does not include photon data, HEATR returns
only the first sum. This is equivalent to assuming that all photon energy is
deposited locally, consistent with the fact that there will be no contribution to
the photon transport source from this material. The same result can be forced
by using the local parameter (see “User Input”, Section 6.7, below).
Discrete photon yields and energies are obtained from MF=12 or 13. Con-
tinuum photon data are obtained from MF=15, and the average photon energy
and Eγ2 are computed. For radiative capture, the photon term becomes
!
E Eγ2
Eγ σγ = E+Q− + Yγ σγ , (171)
A + 1 2(A + 1)mc2

[9] where Yγ is the capture photon yield from MF=12. This corrects the capture
contribution from Eq. 167 by conservation of momentum. For other reactions,
Eq. 167 is sufficient, and the product of E γ , Yγ , and σγ is subtracted from the
neutron contribution.

6.3.2 The special case of fission

The partial KERMA for fission is a special case due to the particular problems
with obtaining the Q-value for fission. First, the fission Q-value given in the C1
field of MF=3 includes delayed neutron and gamma contributions that we need
to exclude, and second, the Q-value for fission is energy dependent.
As a result, the KERMA for fission will be calculated differently when com-
pared to the other reactions which use Eq. 167 as is. Theoretically speaking,
there is no difference with Eq. 167 as we will show here.
Energy dependent fission energy release and its components are given in the
MT=458 section of MF=1 on the ENDF file. This section of the ENDF file
defines the following components to the fission energy release:

122 NJOY2016
6 HEATR LA-UR-17-20093

• Qk : the kinetic energy of the fission fragments

• Qn,p and Qn,d : the kinetic energy of the prompt and delayed fission neu-
trons
• Qγ,p and Qγ,d : the energy of the prompt and delayed gamma rays
• Qβ : the energy of the delayed beta radiation
• Qν : the energy carried away by the neutrinos
With these components, we can now define the total energy release from
fission Qt , the total energy release from fission excluding neutrinos Qr and the
total prompt energy release from fission Qp as as follows:

Qt (E) = Qk (E) + Qn,p + Qn,d + Qγ,p + Qγ,d + Qβ + Qν , (172)

Qr (E) = Qt (E) − Qν , (173)

Qp (E) = Qr (E) − Qn,d − Qγ,d − Qβ = Qk (E) + Qn,p + Qγ,p . (174)

Using these fission energy release components, we can define the fission reac-
tion Q-value (i.e. the energy released through the fission reaction) as the prompt
fission energy release minus the incident neutron energy:

Q(E) = Qp (E) − E = Qk (E) + Qn,p + Qγ,p − E . (175)

It should be noted that we have chosen to ignore the energy dependence of

delayed beta and gamma emission because we don’t yet treat it in subsequent
codes. However, the impact of such an approximation is somewhat limited due
to the amount of energy involved. For example, for U235 the value of Qk is
roughly 169 MeV at 1e-5 eV while the sum of Qn,d , Qγ,d and Qβ is roughly 12
MeV at 1e-5 eV.
For the calculation of the fission KERMA factor, we also need to know the
energy of the outgoing neutrons (i.e. E from Eq. 167). Because we are consider-
ing the prompt energy release only, this is simply equal to the prompt neutron
energy release Qn,p .
As a result, the partial fission KERMA factor kfn will be given by:
h i h i
kfn (E) = E + Q(E) − E(E) σf (E) = Qk (E) + Qγ,p (E) σf (E) . (176)

The fission KERMA is thus equal to the fission cross section times the sum of
the kinetic energy of the fission products and the prompt gamma energy release.

NJOY2016 123
LA-UR-17-20093 6 HEATR

This value will then be used in Eq. 166 to calculate the total KERMA.
In some cases it is possible that a fissile nuclide does not have an MT458
section. In this case, Eq. 167 will be used directly as follows:
h i
kfn (E) = E + Q(E) − ν(E)E(E) σf (E) (177)

where the fission Q-value is approximated using the thermal point energy de-
pendencies defined for MT458:

Q(E) = QENDF − 8070000 (ν(E) − ν(0)) + 0.307E (178)

In this equation, QENDF is the reaction Q-value for fission as given in MF3.

6.4 Kinematic Limits

As an option provided mainly as an aid to evaluators, HEATR will compute the
kinematic maximum and minimum KERMA coefficients and compare them with
the energy-balance results. The formulas are as follows. For elastic scattering
(MT=2), the expected recoil energy is

2AE
ER = (1 − f1 ) . (179)
(A + 1)2
For discrete-inelastic scattering (MT=51-90), the photon momentum is ne-
glected to obtain
" r #
2AE (A + 1)Eγ Eγ
ER = 1 − f1 1− − , (180)
(A + 1)2 AE A+1

where Eγ =−C2 from MF=3. For continuum inelastic scattering (MT=91), sec-
ondary neutrons are assumed to be isotropic in the laboratory system (LAB)
giving

E − En
ER = , (181)
A
and

(A − 1)E − (A + 1)E n
Eγ = , (182)
A

where E γ is the average photon energy expected for this representation. For
radiative capture (MT=102),

124 NJOY2016
6 HEATR LA-UR-17-20093

E
ER = + EK (183)
A+1
and

AE
Eγ = Q + − EK , (184)
A+1
where
 2   
1 AE 1 AE
EK = +Q 1− +Q , (185)
2MR c2 A + 1 MR c2 A + 1
with
MR c2 = (939.565 × 106 )(A + 1) − Q (186)

being the mass energy in eV. The value of this constant is actually computed
from fundamental constants in NJOY2016.
For two-body scattering followed by particle emission (MT=51-91, LR flag
set), a minimum and maximum can be defined:

′
(ER + Ex )min = E R , and (187)
′
(ER + Ex )max = E R + Q + (Eγ )max , (188)

where E R is the value from Eq. 180 or (181), Q is the C2 field from File 3, and
′ is
(Eγ )max is the negative of the C2 field from File 3. In these equations, ER
the recoil energy and Ex is the energy of the charged product. For absorption
followed by particle emission (MT=103-120),

E
(ER + Ex )min = , (189)
A+1−x
A−x
(Eγ )max = Q+ E , and (190)
A+1−x
(ER + Ex )max = E+Q , (191)

where Q is the C2 field from MF=3 and x is the particle mass ratio (x=1 gives
a minimum for all reactions). For (n,2n) reactions,

NJOY2016 125
LA-UR-17-20093 6 HEATR

(ER )min = 0 , and (192)

E + En
(ER )max = , (193)
A−1

and for (n,3n) reactions,

(ER )min = 0 , and (194)

E + 2E n
(ER )max = . (195)
A−2

For both (n,2n) and (n,3n), if (ER )max is greater than ER , it is set equal to ER .
In addition, these formulas are not used for A<10; (ER )max is set to ER . For
other neutron continuum scattering reactions (MT=22-45),

(ER + Ex )min = 0 , and (196)

(ER + Ex )max = E + Q − E n , (197)

where Q is the C2 field from File 3. Finally, for fission (MT=18-21, 38), the
limits are

1
(ER )min = E + Q − E n − 15×106 eV , and (198)
2
(ER )max = E + Q − E n , (199)

where Q is the prompt fission Q-value less neutrinos. It is determined by taking

the total (less neutrinos) value from File 3 and subtracting the delayed energy
obtained from MF=1/MT=458.
These values are intended to be very conservative. Note that EK is only
significant at very low neutron energy. In order to reduce unimportant error
messages, a tolerance band is applied to the above limits. If all checks are
satisfied, the resulting KERMA coefficients should give good local heating results
even when 99.8% of the photons escape the local region. More information on
using the kinematic checks to diagnose energy-balance problems in evaluations
will be found in “Diagnosing Energy-Balance Problems”, Section 6.9.
The upper kinematic limit can also be written out to the output tape as

126 NJOY2016
6 HEATR LA-UR-17-20093

MT=443 if desired. It is similar to the KERMA factors generated by the MACK

code[30], and it is sometimes preferable to the energy-balance KERMA for cal-
culating local heating for evaluations with severe energy-balance problems. The
kinematic value in MT=443 is useful for plots (see the examples in this report).

6.5 Computation of Damage Energy

The formulas used for calculating damage energy are derived from the same
sources as the heating formulas given above, except in this case, the effects of
scattering angle do not result in simple factors like f1 because the Robinson
partition function is not linear. Instead, it is calculated as follows:

ER
P (E) = , (200)
1 + FL (3.4008ϵ1/6 + 0.40244ϵ3/4 + ϵ)

if ER ≥ 25.0 eV, and zero otherwise. In Eq. 200, ER is the primary recoil energy,

ER
ϵ = , (201)
EL
2/3 1/2
 
2/3
EL = 30.724ZR ZL ZR + ZL (AR + AL )/AL , and (202)
2/3 1/2
0.0793ZR ZL (AR + AL )3/2
FL = , (203)
2/3 3/4 3/2 1/2
 
2/3
ZR + ZL AR AL

and Zi and Ai refer to the charge and atomic number of the lattice nuclei (L)
and the recoil nuclei (R). The function behaves like ER at low recoil energies
and then levels out at higher energies. Therefore, the damage-energy production
cross section is always less than the heat production cross section. See Fig. 9 for
examples.
For elastic and two-body discrete-level inelastic scattering,

AE  
ER (E, µ) = 1 − 2Rµ + R 2
, (204)
(A + 1)2

where the “effective mass” is given by

r
(A + 1)(−Q)
R= 1− , (205)
AE
and µ is the CM scattering cosine. The damage energy production cross section

NJOY2016 127
LA-UR-17-20093 6 HEATR

is then obtained from

Z 1
D(E) = σ(E) f (E, µ) P (ER [E, µ]) dµ , (206)
−1

where f is the angular distribution from the ENDF/B File 4. This integration is
performed with a 20-point Gauss-Legendre quadrature. Discrete-level reactions
with LR flags to indicate, for example, (n,n′ )α reactions, are treated in the same
way at present. The additional emitted particles are ignored.
Continuum reactions like (n,n′ ) give a recoil spectrum

1 √ 
ER (E, E ′ , µ) = E − 2 EE ′ µ + E ′ , (207)
A
where E ′ is the secondary neutron energy, µ is the laboratory cosine, and the
photon momentum has been neglected. The damage becomes

Z ∞ Z 1
′
D(E) = σ(E) dE dµ f (E, µ) g(E, E ′ ) P (ER [E, E ′ , µ]) , (208)
0 −1

where g is the secondary energy distribution from File 5. In the code, the
angular distribution is defaulted to isotropic, and a 4-point Gaussian quadrature
is used for the angular integration. For analytic representations of g, an adaptive
integration to 5% accuracy is used for E ′ ; for tabulated File 5 data, a trapezoidal
integration is performed using the energy grid of the file. The same procedure
is used for (n,2n), (n,3n), etc., but it is not realistic for reactions like (n,n′ p)
or (n,n′ α). The neutron in these types of reactions can get out of the nucleus
quite easily; thus, much of the energy available to secondary particles is typically
carried away by the charged particles[49]. HEATR treats these reactions in the
same way as (n,p) or (n,α).
The recoil for radiative capture must include the momentum of the emitted
photons below 25 – 100 keV giving

r s
E E Eγ2 Eγ2
ER = −2 cos ϕ + , (209)
A+1 A + 1 2(A + 1)mc2 2(A + 1)mc2

where ϕ is the angle between the incident neutron direction and emitted photon
direction. If subsequent photons are emitted in a cascade, each one will add
an additional term of Eγ2 and an additional angle. A complete averaging of
Eq. 209 with respect to P (ER ) would be very difficult and would require angular

128 NJOY2016
6 HEATR LA-UR-17-20093

correlations not present in ENDF/B evaluations. However, damage calculations

are still fairly crude, and an estimate for the damage obtained by treating the
neutron “kick” and all the photon kicks independently should give a reasonable
upper limit because
!
Z 1  
Eγ2
E X
D(ER ) d cos ϕ ≤ D + D . (210)
−1 A+1 γ
2MR c2

The actual formula used in the code is

   2 !
E 1 AE
D(E) = D +D +Q
A+1 2MR c2 A + 1
! 2 !
Eγ2

X 1 AE
+ D −D +Q , (211)
γ
2MR c2 2MR c2 A + 1

where the first line is computed in the neutron section, and the second line is
computed in the photon section. This form also provides a reasonable default
when no photons are given.
Finally, for the (n,particle) reactions, the primary recoil is given by

1  ∗ p 
ER = ∗
E − 2 aE Ea cos ϕ + aEa , (212)
A+1

where a is the mass ratio of the emitted particle to the neutron, E ∗ is given by

A+1−a
E∗ = E, (213)
A+1

and the particle energy Ea is approximated as being equal to the smaller of the
available energy

AE
Q+ , (214)
A+1

or the Coulomb barrier energy

1.029 × 106 zZ
in eV , (215)
a1/3 + A1/3

NJOY2016 129
LA-UR-17-20093 6 HEATR

106
total
elastic
105 inelastic
absorption

104
Damage (eV-barns)

103

102

101

0
10

10-1
101 102 103 104 105 106 107
Energy (eV)

Figure 10: Components of radiation damage energy production for 27 Al from ENDF/B-
VII.0. Note that capture dominates at very low energies, then elastic dominates,
and finally inelastic begins to contribute at very high energies.

where z is the charge of the emitted particle and Z is the charge of the target.
A more reasonable distribution would be desirable[49], but this one has the
advantage of eliminating an integration, and most results are dominated by the
kick imparted by the incident neutron anyway. The angular distribution for
the emitted particle is taken as isotropic in the lab. At high incident energies,
direct interaction processes would be expected to give rise to a forward-peaked
distribution, thereby reducing the damage. However, the importance of this
effect is also reduced by the dominance of the neutron kick.
Fig. 10 gives a typical result for a damage energy production calculation,
showing the separate contributions of elastic, inelastic, and absorption processes.

6.6 Heating and Damage from File 6

A number of the evaluations in ENDF/B-VI and later include complete energy-
angle distributions for all of the particles produced by a reaction, including
the residual nucleus. In these cases, HEATR can compute the contributions to
KERMA by calculating the average energy in the spectrum of each outgoing
charged particle or residual nucleus and using Eq. 161.
A fully-populated section of File 6 contains subsections for all of the particles

130 NJOY2016
6 HEATR LA-UR-17-20093

and photons produced by the reaction, including the recoil nucleus. There are a
number of different schemes used to represent the energy-angle distributions for
these outgoing particles. The most important ones for HEATR follow:

• No distribution. In this case, the subsection is inadequate for use in heating

and damage calculations. A warning message is issued.
• Two-body angular distribution. These are basically the same as distributions
in File 4.
• Recoil distribution. This particle is a recoil nucleus from a two-body re-
action. Its angular distribution is assumed to be the complement of the
angular distribution for the first subsection in this section.
• CM Kalbach distribution. This format is often used by LANL evaluations,
and transformation to the laboratory frame is required. The looping order
for the data is E, E ′ , µ.
• LAB Legendre distribution. This format is used in most of the ORNL
evaluations for ENDF/B-VI. It is already in the laboratory frame, and the
angular information can be simply ignored.
• LAB angle-energy distribution. This format is used for the 9 Be evaluation
of ENDF/B-VI by LLNL. The looping order is E, µ, E ′ .
The normal procedure is to loop through all of these subsections. The subsections
producing neutrons are processed to be used in a total energy check, but they
contribute nothing to the heating or to the damage. Subsections describing
charged particles and residual nuclei are processed into heating and damage
contributions. Finally, the photon subsection is processed for the photon energy
check and the total energy check, even though it does not affect either heating
or damage. Any remaining difference between the eV-barns available for the
reaction and the eV-barns carried away by the neutrons, photons, particles, and
recoil is added into the heating to help preserve the total energy deposition in
the spirit of the energy-balance method.
For “two-body” sections, the emitted particle energy is given by

′ A′ E  
E = 2
1 + 2Rµ + R , (216)
A+1

where
r
A(A + 1 − A′ )
R= , (217)
A′
and A′ is the ratio of the mass of the outgoing particle to that of the incident

NJOY2016 131
LA-UR-17-20093 6 HEATR

particle. The heating is obtained by doing a simple integral over µ, and the
damage is computed using the integral over µ given in Eq. 206. In both cases,
the integrals are performed using either a 20-point Gauss-Legendre quadrature
(for Legendre representations) or a trapezoidal integration (for tabulated data).
For “recoil” sections, the code backs up to the particle distribution and cal-
culates the recoil using the same method described above with the sign of µ
changed.
For laboratory distributions that use the E, E ′ , µ ordering, the angular part
can be ignored, and the heating and damage become
Z
K(E) = g(E→E ′ )E ′ dE ′ , (218)

and
Z
D(E) = g(E→E ′ )P (E ′ ) dE ′ , (219)

where g(E→E ′ ) is the angle-integrated energy distribution from File 6, and

P (E ′ ) is the damage partition function. Trapezoidal integration is used for the
continuum, and the integrand is simply added into the sum for the delta functions
(if any).
Heating for subsections that use the ordering E, µ, E ′ is computed using the
formula
Z Z 
′ ′ ′
K(E) = g(E→E , µ)E dE dµ , (220)

where an inner integral is performed using trapezoidal integration for each value
in the µ grid. The results are then used in a second trapezoidal integration over
µ. The damage integral is performed at the same time in a parallel manner.
The problem is somewhat more difficult for subsections represented in the
center-of-mass frame. The definitions for K(E) and D(E) are the same as those
given above, except that the quantity g(E→E ′ ) has to be generated in the lab
system. The methods used to do the transformation are basically the same in
HEATR and GROUPR. The first step is to set up an adaptive integration over
E ′ . The first value needed to prime the stack is obtained by calling h6cm with
E ′ =0. It returns the corresponding value of g in the lab system and a value for
epnext. The second value for the stack is computed for E=epnext. The routine
then subdivides this interval until 2% convergence is achieved, accumulating the
contributions to the heating and damage integrals as it goes. It then moves up

132 NJOY2016
6 HEATR LA-UR-17-20093

to a new panel. This process continues until the entire range of E ′ has been
covered.
The key to this process is h6cm. As described in more detail in the GROUPR
chapter of this manual, it performs integrals of the form
Z +1
gL (E→EL′ ) = gC (E→EC′ , µC ) J dµL , (221)
µmin

where L and C denote the laboratory and center-of-mass systems, respectively,

and J is the Jacobian for the transformation. The contours in the EC′ ,µC frame
that are used for these integrals have constant EL′ . The limiting cosine, µmin ,
depends on kinematic factors and the maximum possible value for EC′ in the File
6 tabulation.
The ENDF/B-VII library contains a few abbreviated versions of File 6 that
contain an energy-angle distribution for neutron emission, but no recoil or photon
data. In order to get semi-reasonable results for both heating and damage for
such cases, HEATR applies a “one-particle recoil approximation,” where the first
particle emitted is assumed to induce all the recoil. There are also some cases
where capture photons are described in MF=6/MT=102 with no corresponding
recoil data. Here, the recoil can be added using the same logic described above for
capture represented using File 15. The difference between the eV-barns available
for the reaction and the energy accounted for by the emitted neutrons, photons,
particles, and the approximated recoil is added into the heating in order to
preserve the total heating in the spirit of the energy-balance method.

6.7 User Input

The input instructions that follow were reproduced from the comment cards in
the current version of HEATR.

!---input specifications (free format)------------------------

!
! card 1
! nendf unit for endf tape
! nin unit for input pendf tape
! nout unit for output pendf tape
! nplot unit for graphical check output
! card 2
! matd material to be processed
! npk number of partial kermas desired (default=0)

NJOY2016 133
LA-UR-17-20093 6 HEATR

! nqa number of user q values (default=0)

! ntemp number of temperatures to process
! (default=0, meaning all on pendf)
! local 0/1=gamma rays transported/deposited locally
! (default=0)
! iprint print (0 min, 1 max, 2 check) (default=0)
! ed displacement energy for damage
! (default from built-in table)
! card 3 for npk gt 0 only
! mtk mt numbers for partial kermas desired
! total (mt301) will be provided automatically.
! partial kerma for reaction mt is mt+300
! and may not be properly defined unless
! a gamma file for mt is on endf tape.
! special values allowed--
! 303 non-elastic (all but mt2)
! 304 inelastic (mt51 thru 91)
! 318 fission (mt18 or mt19, 20, 21, 38)
! 401 disappearance (mt102 thru 120)
! 442 total ev-barns
! 443 total kinematic kerma (high limit)
! damage energy production values--
! 444 total
! 445 elastic (mt2)
! 446 inelastic (mt51 thru 91)
! 447 disappearance (mt102 thru 120)
! cards 4 and 5 for nqa gt 0 only
! card 4
! mta mt numbers for users q values
! card 5
! qa user specified q values (ev)
! (if qa.ge.99.e6, read in variable qbar
! for this reaction)
! card 5a variable qbar (for reactions with qa flag only)
! qbar tab1 record giving qbar versus e (1000 words max)
!
!----------------------------------------------------------------

Card 1 specifies the input and output units for HEATR. They are all ENDF-
type files. The input PENDF file has normally been through RECONR and
BROADR, but it is possible to run HEATR directly on an ENDF file in order

134 NJOY2016
6 HEATR LA-UR-17-20093

to do kinematic checks. In this case, the results in the resonance range should
be ignored. Defining nplot will produce a file of input for the VIEWR module
containing detailed energy-balance test results. This option should only be used
together with iprint=2.
On Card 2, the default value for npk is zero, which instructs the code to
process the energy-balance total KERMA (MT=301) only. Most often, the
user will also want to include MT=443 and MT=444 (npk=2). The kinematic
KERMA computed when MT=443 is requested is very useful for judging the
energy-balance consistency of an evaluation (see the subsection on “Diagnosing
Energy-Balance Problems”, Section 6.9, below). It can also be used instead of the
energy-balance value in MT=301 when local heating effects are important and
the evaluation scores poorly in an energy-balance check. Damage energy pro-
duction cross sections (MT=444) should be computed for important structural
materials; this expensive calculation can be omitted for other materials.
When kinematic checks are desired, a number of additional npk values can
be included. They can be determined by checking the evaluation to see what
partial KERMA factors are well defined. For old-style evaluations that do not
use File 6, look for the MT values used in Files 12 and 13. Many evaluations use
only MT=3 and MT=102 (or 3, 18, and 102 for fissionable materials); in these
cases, the only mtk values that make sense are 302, 303, and 402 (or 302, 303,
318, and 402 for fissionable materials). Caution: in many evaluations, MT=102
is used at low energies and taken to zero at some breakpoint. MT=3 is used
at higher energies. In these evaluations, the partial KERMA MT=402 does not
make sense above the breakpoint, and MT=3 does not make sense below it.
More complicated photon-production evaluations may include MT=4 and/or
discrete-photon data in MT=51-90. In these cases, the user can request mtk=304.
The same kind of energy-range restriction discussed for MT=102 can occur for
the inelastic contributions. Other evaluations give additional partial reactions
that can be used to check the photon production and energy-balance consistency
of an evaluation in detail. HEATR can handle 6 additional reactions at a time.
Multiple runs may be necessary in complex cases.
Note that several special mtk values are provided for the components of the
damage-energy production cross section. They were used to prepare Fig. 10, and
may be of interest to specialists, but they are not needed for most libraries.
In a few cases in the past, it has been necessary to change the Q-values that
are normally retrieved from the ENDF tape. In addition, it is sometimes neces-
sary to replace the single Q-value supplied in MF=3 with an energy-dependent

NJOY2016 135
LA-UR-17-20093 6 HEATR

Q function for an element. One example of the former occurred for 16 O for
ENDF/B-V. The first inelastic level (MT=51) decays by pair production rather
than the more normal mode of photon emission. In order to get the correct
heating, it was necessary to change the Q-value by giving Card 4 and Card 5 as
follows:

51
-5.0294e6

That is, the Q-value is increased by twice the electron energy of 0.511 MeV.
Another example is the sequential (n,2n) reaction for 9 Be in ENDF/B-V. It is
necessary to include 4 changes to the Q-values:

46 47 48 49/
-1.6651e6 -1.6651e6 -1.6651e6 -1.6651e6/

The next example illustrates using energy-dependent Q-values for elemental ti-
tanium. Set nqa equal to 3 and give the following values on Cards 4, 5, and
5a:

16 103 107/
99e6 99e6 99e6/
0. 0. 0 0 1 8
8 2
8.0e6 -8.14e6 9.0e6 -8.14e6 1.1e7 -8.38e6
1.2e7 -8.74e6 1.3e7 -1.03e7 1.4e7 -1.091e7
1.5e7 -1.11e7 2.0e7 -1.125e7/
0. 0. 0 0 1 9
9 2
1.0e-5 1.82e5 4.0e6 1.82e5 5.0e6 -1.19e6
6.0e6 -2.01e6 7.0e6 -2.20e6 8.0e6 -2.27e6
1.4e7 -2.35e6 1.7e6 -2.43e6 2.0e7 -2.37e6/
0. 0. 0 0 1 9
9 2
1.0e-5 2.182e6 6.0e6 2.182e6 7.0e6 2.10e6
8.0e6 -3.11e5 9.0e6 -9.90e5 1.0e7 -1.20e6

136 NJOY2016
6 HEATR LA-UR-17-20093

1.1e7 -1.27e6 1.4e6 -1.32e6 2.0e7 -1.48e6/

The next parameter on Card 2 is ntemp. For normal runs, use zero, and all
the temperatures on the input PENDF tape will be processed. For kinematic
check runs, use ntemp=1. The local parameter suppresses the processing of
the photon-production files, if any. The photon energy appears in the KERMA
factors as if the photons had very short range. A useful way to use the iprint
parameter is to set it to zero for normal runs, which produce heating and damage
values at all temperatures, and to use iprint=2 for the energy-balance check
run, which is performed for the first temperature on nin only.
Card 3 gives the partial KERMA and damage selection MT numbers. Note
that the user does not include MT=301 in this list. It is always inserted as the
first value automatically. Giving MT=301 in this list will cause an informative
message to be issued.
Cards 4, 5, and 5a give the user’s changes to the ENDF Q values. The way
in which to use these cards was described in connection with nqa on Card 2.

6.8 Reading HEATR Output

When full output and/or kinematic checks have been requested, HEATR loops
through the reactions found in Files 3, 12, and 13. For each reaction, it prints
out information about the energies, yields, cross sections, and contributions to
heating. The energy grid used is a subset of the PENDF grid. At present,
decade steps are used below 1 eV, factor-of-two steps are used from 1 eV to 100
keV, quarter-lethargy steps are used above 100 keV, and approximately 1 MeV
steps are used above 2 MeV. An example of this printout for elastic scattering
in ENDF/B-VII.0 27 Al is shown below:

neutron heating for mt 2 q0 = 0.0000E+00 q = 0.0000E+00

e ebar ... xsec heating damage
1.0000E-05 9.3052E-06 ... 1.5694E+01 1.0903E-05 0.0000E+00
1.0000E-04 9.3052E-05 ... 5.1179E+00 3.5557E-05 0.0000E+00
1.0000E-03 9.3052E-04 ... 2.0644E+00 1.4342E-04 0.0000E+00
1.0000E-02 9.3052E-03 ... 1.4925E+00 1.0369E-03 0.0000E+00
1.0000E-01 9.3052E-02 ... 1.4318E+00 9.9474E-03 0.0000E+00
...
1.0000E+03 9.3052E+02 ... 1.3662E+00 9.4914E+01 7.7171E+01
2.0000E+03 1.8610E+03 ... 1.3196E+00 1.8337E+02 1.5115E+02

NJOY2016 137
LA-UR-17-20093 6 HEATR

5.0000E+03 4.6526E+03 ... 1.2130E+00 4.2136E+02 3.3998E+02

1.0000E+04 9.3052E+03 ... 1.0367E+00 7.2028E+02 5.6725E+02
2.0000E+04 1.8610E+04 ... 6.6204E-01 9.1991E+02 7.0347E+02
5.0000E+04 4.6526E+04 ... 2.3220E+00 8.0660E+03 5.8732E+03
1.0000E+05 9.3052E+04 ... 5.2976E+00 3.6805E+04 2.5521E+04
...
1.0000E+07 9.7233E+06 ... 7.4942E-01 2.0738E+05 4.5689E+04
1.1000E+07 1.0699E+07 ... 7.4953E-01 2.2576E+05 4.7347E+04
1.2000E+07 1.1682E+07 ... 7.6363E-01 2.4286E+05 4.9202E+04
1.3000E+07 1.2673E+07 ... 7.6329E-01 2.4980E+05 4.9559E+04
1.4000E+07 1.3670E+07 ... 7.7918E-01 2.5737E+05 5.0566E+04
1.5000E+07 1.4671E+07 ... 7.9297E-01 2.6088E+05 5.1189E+04
1.6000E+07 1.5676E+07 ... 8.1414E-01 2.6347E+05 5.1973E+04
1.7000E+07 1.6681E+07 ... 8.3429E-01 2.6577E+05 5.2680E+04
1.8000E+07 1.7689E+07 ... 8.4524E-01 2.6246E+05 5.2747E+04
1.9000E+07 1.8695E+07 ... 8.7093E-01 2.6568E+05 5.3787E+04
2.0000E+07 1.9703E+07 ... 8.9326E-01 2.6547E+05 5.4573E+04
...
1.4000E+08 1.3985E+08 ... 2.9730E-01 4.3523E+04 1.4994E+04
1.4600E+08 1.4585E+08 ... 2.7580E-01 4.1851E+04 1.3890E+04
1.5000E+08 1.4984E+08 ... 2.6510E-01 4.1161E+04 1.3339E+04

Note the identification and Q information printed on the first line; q is the ENDF
Q-value from File 3, and q0 is the corresponding mass-difference Q-value needed
for Eq. 162. The ebar, yield (which was replaced by “...” to make this listing
fit better), and xsec columns contain E n , Y , and σ, respectively. The heating
column is just (E+Q−Y E n )σ. The results are similar for discrete inelastic
levels represented using File 4. The heating due to the associated photons will
be subtracted later while MF=12 or MF=13 is being processed. However, if
an LR flag is set, the residual nucleus from the (n,n′ ) reaction breaks up by
emitting additional particles. This extra breakup energy changes the q0 value.
An example of such a section for 27 Al(n,n )p
25 from ENDF/B-V follows:

neutron heating for mt 75 q0 = -8.2710e+06 q = -1.0750e+07

e ebar yield xsec heating
1.2000e+07 7.8653e+05 1.0000e+00 8.2242e-02 2.4199e+05
1.3000e+07 1.7116e+06 1.0000e+00 8.0121e-02 2.4176e+05
1.4000e+07 2.6427e+06 1.0000e+00 5.9282e-02 1.8296e+05
1.5000e+07 3.5864e+06 1.0000e+00 4.1834e-02 1.3147e+05

138 NJOY2016
6 HEATR LA-UR-17-20093

1.6000e+07 4.5096e+06 1.0000e+00 2.8880e-02 9.2977e+04

1.7000e+07 5.4335e+06 1.0000e+00 1.9867e-02 6.5472e+04
1.8000e+07 6.3848e+06 1.0000e+00 1.3677e-02 4.5739e+04
1.9000e+07 7.2944e+06 1.0000e+00 9.4771e-03 3.2550e+04
2.0000e+07 8.2479e+06 1.0000e+00 6.6142e-03 2.3025e+04

Starting with ENDF/B-VI, discrete-inelastic sections may also be given in File

6. Such sections contain their own photon production data, and the heating
column will represent the entire recoil energy as in Eq. 180. (See below for
detailed discussion of ENDF/B-VI output.)
For continuum reactions that use MF=4 and MF=5, such as (n,n′ ) or (n,2n),
the neutron part of the display looks like this:

neutron heating for mt 16 q0 = -1.3057e+07 q = -1.3057e+07

e ebar yield xsec heating
1.4000e+07 1.9960e+05 2.0000e+00 2.4000e-02 1.3051e+04
1.5000e+07 6.6850e+05 2.0000e+00 1.2320e-01 7.4659e+04
1.6000e+07 1.0855e+06 2.0000e+00 2.0710e-01 1.5987e+05
1.7000e+07 1.4308e+06 2.0000e+00 2.6510e-01 2.8667e+05
1.8000e+07 1.6379e+06 2.0000e+00 3.0300e-01 5.0518e+05
1.9000e+07 1.7659e+06 2.0000e+00 3.3000e-01 7.9567e+05
2.0000e+07 1.8755e+06 2.0000e+00 3.5000e-01 1.1172e+06

Once again, the photon effects will be subtracted later.

Absorption reactions such as (n,γ) or (n,p), lead to similar displays, but the
particle ebar columns will always be set to zero (no emitted neutrons). An
example follows:

neutron heating for mt103 q0 = -1.8278e+06 q = -1.8278e+06

e ebar yield xsec heating
2.5000E+06 0.0000E+00 1.0000E+00 3.2800E-05 2.2048E+01
3.0000E+06 0.0000E+00 1.0000E+00 1.3300E-03 1.5590E+03
3.5000E+06 0.0000E+00 1.0000E+00 1.0100E-02 1.6889E+04
4.0000E+06 0.0000E+00 1.0000E+00 6.9667E-03 1.5133E+04
4.5000E+06 0.0000E+00 1.0000E+00 1.7000E-02 4.5427E+04
5.0000E+06 0.0000E+00 1.0000E+00 2.3300E-02 7.3912E+04
...
1.7000E+07 0.0000E+00 1.0000E+00 5.5200E-02 8.3751E+05

NJOY2016 139
LA-UR-17-20093 6 HEATR

1.8000E+07 0.0000E+00 1.0000E+00 4.7800E-02 7.7303E+05

1.9000E+07 0.0000E+00 1.0000E+00 4.0200E-02 6.9032E+05
2.0000E+07 0.0000E+00 1.0000E+00 3.2200E-02 5.8514E+05

If File 6 is present (which happens for evaluations in ENDF-6 format only,

such as the evaluations in ENDFB-VII), each reaction will be divided into sub-
sections, one for each emitted particle. The neutron subsections are displayed
as part of the energy-balance checks, but they do not contribute to KERMA
or damage. The subsection for each charged particle or residual nucleus will
give the incident energy, average energy for the emitted particle, cross section,
heating contribution, and (optionally) damage contribution as follows:

file six heating for mt 28, particle = 1 q = -8.2721E+06

e ebar yield xsec heating
9.0000E+06 1.2303E+05 1.0000E+00 1.0385E-03 0.0000E+00
1.0000E+07 4.6746E+05 1.0000E+00 1.3526E-02 0.0000E+00
...
1.8000E+07 3.1862E+06 1.0000E+00 3.7721E-01 0.0000E+00
1.9000E+07 3.4535E+06 1.0000E+00 3.7577E-01 0.0000E+00
2.0000E+07 3.7207E+06 1.0000E+00 3.7434E-01 0.0000E+00

file six heating for mt 28, particle = 1001 q = -8.2721E+06

e ebar yield xsec heating
9.0000E+06 4.5909E+05 1.0000E+00 1.0385E-03 4.7677E+02
1.0000E+07 8.9616E+05 1.0000E+00 1.3526E-02 1.2121E+04
...
1.8000E+07 3.5193E+06 1.0000E+00 3.7721E-01 1.3275E+06
1.9000E+07 3.8257E+06 1.0000E+00 3.7577E-01 1.4376E+06
2.0000E+07 4.1321E+06 1.0000E+00 3.7434E-01 1.5468E+06

file six heating for mt 28, particle = 12026 q = -8.2721E+06

e ebar yield xsec heating
9.0000E+06 3.2104E+05 1.0000E+00 1.0385E-03 3.3340E+02
1.0000E+07 3.8540E+05 1.0000E+00 1.3526E-02 5.2128E+03
...
1.8000E+07 8.0820E+05 1.0000E+00 3.7721E-01 3.0486E+05
1.9000E+07 8.6147E+05 1.0000E+00 3.7577E-01 3.2372E+05
2.0000E+07 9.1475E+05 1.0000E+00 3.7434E-01 3.4243E+05

file six heating for mt 28, particle = 0 q = -8.2721E+06

140 NJOY2016
6 HEATR LA-UR-17-20093

e ebar yield xsec heating

9.0000E+06 1.8347E+05 2.8104E-06 1.0385E-03 0.0000E+00
ebal -1.8204E+02
1.0000E+07 4.4913E+05 3.0441E-05 1.3526E-02 0.0000E+00
ebal -2.8626E+02
...
1.8000E+07 1.8309E+06 1.2028E+00 3.7721E-01 0.0000E+00
ebal 4.5397E+03
1.9000E+07 1.8856E+06 1.3695E+00 3.7577E-01 0.0000E+00
ebal 1.8336E+03
2.0000E+07 1.9403E+06 1.5363E+00 3.7434E-01 0.0000E+00
ebal -7.6798E+03

Note that the last subsection in this example was for emitted photons. Pho-
tons do not contribute to the KERMA or damage, but this information is used
to check the total energy conservation for this reaction. The ebal lines show
the difference between the available energy and the sum over all the outgoing
particles. The values should be a small percentage of the total heating. If the
ebal values are too large, there may be an error in the evaluation, or it may be
necessary to refine the energy grids in the distributions. In addition, this photon
production information is needed later for the photon energy check.
After all the sections corresponding to MT numbers in File 3 have been
processed (using the File 4, File 4/5, or File 6 method as appropriate), the photon
production sections in Files 12 and 13 are processed, if present. File 12 data are
usually present for radiative capture (MT=102), at least at low energies. Simple
files normally give a tabulated photon spectrum. The display gives the average
energy for this spectrum in the ebar column and the negative contribution to
the heating computed with Eq. 171 in the heating column. The edam column
contains the Eγ2 term needed to compute the photon contribution to the damage,
which is given in the damage column. See Eq. 211. The display also has an extra
line for each incident energy containing the percent error “–- pc” between the
total photon energy as computed from File 12 and the value E+Q−E/(A+1)
computed from File 3. As discussed above, HEATR does not guarantee energy
balance in large systems if this error occurs. The following example shows some
large errors due to mistakes in the ENDF/B-V evaluation for 55 Mn. Two columns
labeled edam and xsec have been removed to show the heating and damage
columns. The text has also been shifted to the left of its normal position to fit
better on the printed page.

NJOY2016 141
LA-UR-17-20093 6 HEATR

photon energy (from yields) mf12, mt102

e ebar/err egam ... yield heating damage
1 continuum gammas
1.0000e-05 4.5088e+06 2.4237e+02 ... 2.4791e+00 -4.8477e+09 5.4975e+04
1.0000e-05 53.7 pc
1.0703e-04 4.5088e+06 2.4237e+02 ... 2.4791e+00 -1.4819e+09 1.6806e+04
1.0703e-04 53.7 pc
1.2520e-03 4.5088e+06 2.4237e+02 ... 2.4791e+00 -4.3347e+08 4.9159e+03
1.2520e-03 53.7 pc
...
1.3571e+04 4.4522e+06 2.3887e+02 ... 2.4836e+00 -7.3869e+03 -1.2406e-01
1.3571e+04 51.8 pc
2.7142e+04 4.3957e+06 2.3536e+02 ... 2.4881e+00 -3.7037e+05 -1.6487e+01
2.7142e+04 49.9 pc
5.4287e+04 4.2826e+06 2.2834e+02 ... 2.4972e+00 -2.5604e+04 -2.5340e+00
5.4287e+04 46.0 pc
1.0858e+05 4.0563e+06 2.1430e+02 ... 2.5154e+00 -1.3327e+04 -2.7289e+00
1.0858e+05 38.3 pc

Many MF=12, MT=102 sections give multiplicities for the production of

discrete photons. In these cases, HEATR prints out data for all of the parts, and
it provides a sum at the end. The balance error is printed with the sum. The
following example shows a case with discrete photons. The last two columns
have been removed (heating, damage), and the text has been compacted and
shifted to the left to fit on the printed page.

photon energy (from yields) mf12, mt102

e ebar/err egam edam xsec yield
1 7.7260E+06 ev gamma
1.0000e-05 7.7260e+06 1.1448e+03 8.9780e+02 1.1677e+01 3.0000e-01
1.1406e-04 7.7260e+06 1.1448e+03 8.9780e+02 3.4574e+00 3.0000e-01
1.1406e-03 7.7260e+06 1.1448e+03 8.9780e+02 1.0934e+00 3.0000e-01
...
2.0000e+07 2.7005e+07 1.1448e+03 8.9780e+02 1.0000e-03 3.0000e-01
2 7.6950e+06 ev gamma
1.0000e-05 7.6950e+06 1.1356e+03 8.9091e+02 1.1677e+01 5.0000e-02
1.1406e-04 7.6950e+06 1.1356e+03 8.9091e+02 3.4574e+00 5.0000e-02
...

142 NJOY2016
6 HEATR LA-UR-17-20093

2.0000e+07 2.6974e+07 1.1356e+03 8.9091e+02 1.0000e-03 5.0000e-02

3 6.8630e+06 ev gamma
1.0000e-05 6.8630e+06 9.0330e+02 7.1515e+02 1.1677e+01 1.2000e-03
1.1406e-04 6.8630e+06 9.0330e+02 7.1515e+02 3.4574e+00 1.2000e-03
1.1406e-03 6.8630e+06 9.0330e+02 7.1515e+02 1.0934e+00 1.2000e-03
...
89 3.1000e+04 eV gamma
1.0000e-05 3.1000e+04 1.8430e-02 0.0000e+00 1.1677e+01 2.8884e-01
1.0000e-05 0.0 pc
1.1406e-04 3.1000e+04 1.8430e-02 0.0000e+00 3.4574e+00 2.8884e-01
1.1406e-04 0.0 pc
1.1406e-03 3.1000e+04 1.8430e-02 0.0000e+00 1.0934e+00 2.8884e-01
1.1406e-03 0.0 pc
1.1912e-02 3.1000e+04 1.8430e-02 0.0000e+00 3.3832e-01 2.8884e-01
1.1912e-02 0.0 pc
...
2.0000e+07 3.1000e+04 1.8430e-02 0.0000e+00 1.0000e-03 2.8884e-01
2.0000e+07 0.0 pc

In this case (27 Al), the capture energy production checks out perfectly for the
sum of all 89 discrete photons.
Other sections using either File 12 or File 13 generate displays similar to the
following:

photon energy (from xsecs) mf13, mt 3

e ebar xsec energy heating
1 continuum gammas
2.0000e+05 3.6753e+06 4.2076e-03 1.5464e+04 -1.5464e+04
4.0500e+05 3.3863e+06 5.2873e-03 1.7904e+04 -1.7904e+04
6.0031e+05 3.1097e+06 6.4478e-03 2.0051e+04 -2.0051e+04
8.0182e+05 2.0089e+06 9.3236e-02 1.8730e+05 -1.8730e+05
1.0000e+06 9.2622e+05 1.7859e-01 1.6541e+05 -1.6541e+05
1.2000e+06 9.6151e+05 2.8329e-01 2.7239e+05 -2.7239e+05
...

Note that the photon Eσ is simply subtracted from the heating column for each
incident energy.
If the partial KERMA mtk=443 was requested in the user’s input, HEATR
will print out a special section that tests the total photon energy production

NJOY2016 143
LA-UR-17-20093 6 HEATR

against the kinematic limits (see Section 6.4 above for the formulas used). An
example follows:

photon energy production check

e ev-barns min max
1.0000e-05 9.0215e+07 9.0187e+07 9.0200e+07
1.1406e-04 2.6712e+07 2.6704e+07 2.6708e+07
1.1406e-03 8.4479e+06 8.4453e+06 8.4466e+06
1.1912e-02 2.6138e+06 2.6130e+06 2.6134e+06
1.2812e-01 7.9895e+05 7.9871e+05 7.9883e+05
1.2812e+00 2.5211e+05 2.5203e+05 2.5207e+05
2.6875e+00 1.7420e+05 1.7415e+05 1.7417e+05
5.5000e+00 1.2186e+05 1.2182e+05 1.2184e+05
1.1406e+01 8.4662e+04 8.4636e+04 8.4648e+04
2.4062e+01 5.8231e+04 5.8213e+04 5.8222e+04
4.9375e+01 4.0614e+04 4.0601e+04 4.0607e+04
1.0000e+02 2.8522e+04 2.8514e+04 2.8518e+04
...
8.0000e+06 3.8972e+06 3.7964e+06 4.4880e+06
9.0000e+06 4.4782e+06 4.4401e+06 5.4986e+06
1.0000e+07 4.9645e+06 5.2078e+06 6.6176e+06
1.1000e+07 5.3712e+06 6.1302e+06 7.8374e+06 ----
1.2000e+07 5.3212e+06 5.8699e+06 8.0618e+06
1.3000e+07 5.0984e+06 5.9333e+06 8.5312e+06 ----
1.4000e+07 4.7415e+06 5.7172e+06 8.6605e+06 ----
1.5000e+07 4.0795e+06 4.7419e+06 8.0409e+06 ----
1.6000e+07 3.2521e+06 3.6806e+06 7.2734e+06 ----
1.7000e+07 2.8079e+06 2.8418e+06 6.6927e+06
1.8000e+07 2.7492e+06 2.2915e+06 6.2850e+06
1.9000e+07 2.9626e+06 1.9674e+06 6.2330e+06
2.0000e+07 3.4419e+06 1.7938e+06 6.1994e+06

The low and high kinematic limits will be the same at low energies where only
kinematics affect the calculations. They may be the same for all energies for
ENDF/B-VII evaluations that provide complete distributions for all outgoing
charged particles and recoil nuclei. Normally, the limits diverge above the thresh-
old for continuum reactions. Note that HEATR marks lines where the computed
value goes more than a little way outside the limits with the symbols ++++ or
––. It is often convenient to extract these numbers from the output listing and

144 NJOY2016
6 HEATR LA-UR-17-20093

plot them (see Fig. 11). Although the energy grid is a little coarse, such plots
can often be useful (see below).
The last part of a full HEATR output listing is a tabulation of the computed
KERMA and damage coefficients on the normal coarse energy grid. Columns
are provided for the total KERMA and for each of the partial KERMA results
requested with mtk values in the user’s input. If kinematic checks were requested,
the check values are written just above and below the corresponding partial
KERMA values. In addition, low and high messages are written just above or
just below the kinematic limits in every column where a significant violation of
the limits occurs. Caution: if summation reactions (MT=3, MT=4) were used
to define the photon production over some parts of the energy range, the partial
KERMA results may not make sense at some energies. For example, consider
the common pattern in ENDF/B-V where MT=102 is used for capture at low
energies, but at higher energies, it is set to zero, and the capture contribution is
included in MT=3 (nonelastic). Clearly, the partial KERMA MT=402 doesn’t
make sense above this breakpoint. The following example shows part of the final
KERMA listing for ENDF/B-V.1 55 Mn. The damage column was removed and
the columns compressed to fit on the printed page.

Figure 11: Example of a plot comparing the total photon energy production for 55 Mn from
ENDF/B-V.1 (dashed) with the kinematic limits (solid).

NJOY2016 145
LA-UR-17-20093 6 HEATR

final kerma factors

e 301 302 303 402 443

min 1.2400e-04 3.7775e-06 1.2022e-04 1.2022e-04

1.0000e-05 4.0068e+05 3.7775e-06 4.0068e+05 4.0068e+05 4.0068e+05
max 3.3820e+05 3.7775e-06 3.3820e+05 3.3820e+05
high high high

...

min 3.2373e+04 2.3497e+04 8.8759e+03 4.1114e+01

6.0031e+05 1.0017e+05 2.3497e+04 7.6673e+04 2.9899e+04 3.2375e+04
max 3.2375e+04 2.3497e+04 8.8781e+03 4.3366e+01
high high high

low low
min 7.3041e+04 5.9403e+04 1.3638e+04 4.4748e+01
8.0182e+05 -2.4355e+04 5.9403e+04 -8.3758e+04 2.4987e+04 7.3043e+04
max 7.3043e+04 5.9403e+04 1.3640e+04 4.6676e+01
high

low low
min 9.8973e+04 7.7075e+04 2.1898e+04 4.7777e+01
1.0000e+06 3.7682e+04 7.7075e+04 -3.9393e+04 2.1917e+04 9.8974e+04
max 9.8974e+04 7.7075e+04 2.1900e+04 4.9509e+01
high

low low
min 1.1397e+05 7.7800e+04 3.6168e+04 4.9760e+01
1.2000e+06 9.5321e+04 7.7800e+04 1.7521e+04 1.9482e+04 1.1397e+05
max 1.1397e+05 7.7800e+04 3.6169e+04 5.1335e+01
high

low low
min 1.4632e+05 1.0192e+05 4.4402e+04 5.3005e+01
1.4000e+06 9.8251e+04 1.0192e+05 -3.6667e+03 1.8208e+04 1.4632e+05
max 1.4632e+05 1.0192e+05 4.4403e+04 5.4511e+01
high

...

146 NJOY2016
6 HEATR LA-UR-17-20093

min 2.3650e+05 1.6907e+05 6.7426e+04 1.7607e+02

1.9000e+07 7.1384e+06 1.6907e+05 6.9693e+06 1.3503e+04 2.5435e+06
max 2.5435e+06 1.6907e+05 2.3744e+06 1.7939e+02
high high high

min 2.4001e+05 1.8261e+05 5.7406e+04 1.4423e+02

2.0000e+07 9.2369e+06 1.8261e+05 9.0543e+06 1.0908e+04 2.8385e+06
max 2.8385e+06 1.8261e+05 2.6559e+06 1.4701e+02
high high high

The following subsection discusses how to analyze the “check” output of

HEATR in order to diagnose energy-balance errors in ENDF-format evaluations.
The examples are drawn from ENDF/B-V testing[43]. In general, results like
these are less likely to occur in modern evaluations.

6.9 Diagnosing Energy-Balance Problems

The analysis should start with MT=102, because if it is wrong, the guarantee of
energy conservation for large systems breaks down. If the display for MF=12,
MT=102 shows messages of the form “–- pc”, there may be a problem. If these
messages only show up at the higher energies, and if the size of the error increases
with energy, it is probable that the evaluator has used a thermal spectrum over
the entire energy range (this is very common). Of course, the total photon energy
production from radiative capture should equal

A
E+Q , (222)
A+1
where the rest of the total energy E+Q is carried away by recoil. If only a
thermal spectrum is given, the E term is being neglected, and errors will normally
appear above about 1 MeV. The E term can be included in evaluations that use
tabulated data by giving E-dependent spectra in File 15; and it can be included
for evaluations that use discrete photons by setting the “primary photon” flags
in File 12 properly. In practice, the capture cross sections above 1 MeV are
often comparatively small due to the 1/v tendency of capture, and the errors
introduced by neglecting the E term can be ignored.
If the MT=102 errors show up at low energies, there is probably an error
in the average photon yield from File 12, in the average energy computed from
File 15, or both. In the 55 Mn case shown above, the yield had been incorrectly

NJOY2016 147
LA-UR-17-20093 6 HEATR

entered. In addition, the spectrum didn’t agree with the experimental data
because the bin boundaries were shifted. Each case must be inspected in detail
to find the problems.
The next common source of energy-balance errors in ENDF files arises from
the representation used for inelastic scattering. Typically, the neutron scattering
is described in detail using up to 40 levels for the (n,n′ ) reaction. However, the
photon production is often described using MF=13/MT=3 or MF=13/MT=4
and rather coarse energy resolution. As a result, it is possible to find photons for
(n,n1 ) being produced for incident neutron energies slightly below the MT=51
threshold! These photons would lead to a spike of negative KERMA factors. A
more common effect of the coarse grid used for photon production is to lead to
an underestimate or overestimate of the photon production by not following the
detailed shape of the inelastic cross section. The HEATR “kinematic KERMA”
is correct in this range since only two-body reactions are active. Therefore, a
plot of MT=301 and MT=443 on the same frame normally shows these effects
in detail. Fig. 12 is an example of such a plot.
Fig. 13 shows both the inelastic cross section from File 3 and the photon pro-
duction cross section from File 13 to demonstrate the mismatch in the energy
grids that contributes to the energy-balance errors. These kinds of errors are
best removed by changing to a representation that uses File 12 to give photon

Figure 12: Comparison of MT=301 with MT=443 for the region of the discrete-inelastic
thresholds for 59 Co from ENDF/B-V.2. Note the large region of negative
KERMA. The best way to remove this kind of problem is by using yields in
File 12, MT=51, 52, 53, . . . to represent the photon production.

148 NJOY2016
6 HEATR LA-UR-17-20093

Figure 13: Plot showing the mismatch between the energy grids used for File 3 and File 13
in the region of the thresholds for discrete-inelastic scattering levels for the case
shown in Fig. 12. The cross and ex symbols show the actual grid energies in the
evaluation.

production yields for the separate reactions MT=51, MT=52, etc. This repre-
sentation makes full use of the File 3 cross sections, and as long as each section
of File 12 conserves energy, the total inelastic reaction is guaranteed to conserve
energy, even at the finest energy resolution.

Figure 14: Typical energy-balance problems between points where balance is satisfied. Dis-
crete photons were used below about 2 MeV, and energy balance is reasonably
good there. The energy points in MF=13 for the continuum part are at 2, 3, and
5 MeV, and the balance is also good at those energies. Clearly, a grid in File 13
that used steps of about 0.25 MeV between 2 and 4 MeV would reduce the size
of the deviations substantially and remove the negative KERMA factors.

NJOY2016 149
LA-UR-17-20093 6 HEATR

A method that is frequently used by evaluators of photon production files is

to select a number of nonelastic photon spectra on a fairly coarse incident-energy
grid using theory or experiment, and then to readjust the photon yield on this
energy grid so as to conserve energy at each grid point. However, the results do
not, in general, conserve energy at intermediate points. If a very coarse energy
grid is used for File 13, quite large deviations between MT=301 and MT=443 can
result. Fig. 14 shows such a case. The solution to this kind of violation of energy
balance is to add intermediate points in Files 13 and 15 until the magnitude of
the deviations is small enough for practical calculations.
Especially large energy-balance errors of this type are caused by interpolating
across the minimum formed by the decreasing capture heating and the increasing
inelastic heating. Fig. 15 shows a dramatic example using a photon energy
production comparison.
For energies above the threshold for continuum reactions like (n,n′ ) or (n,2n),
it is difficult to use the results of the kinematic checks to fix evaluations. The
representation of Eqs. 181 and 182 for continuum inelastic scattering is very
rough. Comparison to other more accurate methods suggests that a CM formula

Figure 15: Computed photon energy production (dashed) compared with the kinematic
value (solid) for 93 Nb from ENDF/B-V. The original File 13 has grid points
at 100 keV and 1 MeV. Interpolating across that wide bin gives a photon produc-
tion rate that is much too large for energies in the vicinity of a few hundred keV.
This will result in a large region of negative heating heating numbers. Since this
is just the region of the peak flux in a fast reactor, niobium-clad regions could
be cooled instead of heated!

150 NJOY2016
6 HEATR LA-UR-17-20093

would be better here[50], even though the ENDF file says “lab.” Most other
reactions give very wide low and high limits. Two exceptions are (n,2n) and
(n,3n). If they dominate the cross section, the kinematic limits will be fairly
close together. In the 14 MeV range, energy errors could be in the photon
data, the neutron data, or both. The best way to eliminate balance errors is to
construct a new evaluation based on up-to-date nuclear model codes.

6.10 Coding Details

The main subroutine is heatr, which is exported by module heatm. It starts by
reading the user’s input and locating the desired material on the PENDF file.
The main loop is over temperature. For each temperature, a check is made to
see if the user provided a value for the damage displacement energy. If not, a
default value is provided. Next, hinit is called to examine the directory. Flags
are set if MF=12 or 13 is present, if MT=18 or 19 is used, and if MT=458 is
present (see mgam, mt19, and mt458). The flags mt103, mt104, mt105, mt106
and mt107 are set if the corresponding particle production levels are present.
The MT numbers used for the levels depend on whether the input file used
version 6 format or one of the earlier formats. For example, mt103 is set if
MT=600-649 is found for ENDF-6 data, or if MT=700-719 is found for earlier
versions. The code also checks to see if the corresponding angular distribution
data are present (see nmiss4). If any are not present, the code will assume they
are isotropic. Note that hinit also collects a list of the File 6 MT numbers in
mt6(i6). For fissionable materials, the delayed fission energy values are retrieved
from MF=1/MT=458, and the correction, qdel, is computed for later use when
calculating the heating from prompt fission.
The next step in hinit is to make a copy of File 6 on a scratch file (if
any sections of File 6 were found). While doing this, it searches through the
subsections for each reaction accumulating the ZA residual remaining after each
particle is given. If it comes to the photons (zap=0), which should be last, and
there is still a ZA residual left, then it concludes that there is no subsection
describing the residual. It loads a value into mt6no(ii6) that is the index to
the subsection that the recoil should have followed had it been present. If it
comes to the end of the subsection list without finding photons and still has a
ZA residual, it sets mt6no(ii6) to nk; that is, the residual missing should follow
the last subsection. In either case, the routine prints out messages about “photon
recoil correction” or “one-particle recoil approximation.”
Finally, hinit makes a standardized copy of the ENDF tape using hconvr,

NJOY2016 151
LA-UR-17-20093 6 HEATR

and it also saves the grid of the total cross section (MT=1) on the loada/finda
scratch file that will be used to accumulate the KERMA factors, damage, and
kinematic checks (if requested). Note also mt303, which tells which of the re-
quested edits is for nonelastic heating, MT=303. This is used later for writing
out the photon energy production check.
Now nheat is called. Its basic function is to loop over the “nonredundant”
reactions in File 3, and to accumulate the corresponding contributions to the
partial heating and partial damage values into the appropriate elements of the c
array on the loada/finda file. Redundant reactions are reactions that duplicate
or include effects that can be obtained from another MT number. They are
determined using a set of if statements just after the entry to the reaction loop
at statement number 105. The structure of the c array depends on whether
kinematic checks are being accumulated or not and whether photon production
files are present. If neither occurs, the structure has npk+1 elements as follows:

Element Contents
1 energy
2 total heating
3 value for first partial
···
npk+1 value for last partial

where npk is the number of partial KERMA or damage values being accumulated,
including the total. If checks are being accumulated, the c array has the following
3*npk+1 elements:

Element Contents
1 energy
2 total heating
3 value for first partial
···
2+npk lower kinematic limit for total
3+npk lower kinematic limit for first partial
···
2+2*npk upper kinematic limit for total
3+2*npk upper kinematic limit for first partial
···
3*npk+1 upper kinematic limit for last partial

If photon production files are present in the evaluation, the total length of the

152 NJOY2016
6 HEATR LA-UR-17-20093

c array increases by the following three words (len is the old length from above
plus three):

Element Contents
len-2 photon capture correction
len-1 total photon eV-barns
len total energy yield for “subtot”

Back inside the loop over nonredundant reactions, subroutine gety1 is ini-
tialized for this reaction. The code checks to see if this section uses File 6 for its
distributions; if so, it arranges to make multiple passes through the reaction’s
energy grid, one pass for each subsection of the MF=6 section, and perhaps one
additional pass to synthesize the missing recoil subsection. It is now possible to
select the appropriate Q-value and particle yield, and to initialize the appropriate
calculational routine. This routine will be sixbar for all reactions described in
File 6, disbar for two-body reactions using File 4 (including charged-particle re-
actions in the 600 or 700 series of MT numbers), conbar for continuum reactions
represented using File 5, and capdam for the neutron disappearance reactions
(MT=102, 103, etc.) and the charged-particle continuum reactions from the 600
or 700 series of MT numbers. The last step before beginning the energy loop for
this reaction is to call indx, which determines which element of the c array is to
receive the heating or damage contribution from this reaction (see below).
The energy loop in nheat goes through statement number 190. For each
energy, finda is called to retrieve the current values for the energy [see c(1)] and
the partial heating and damage values as accumulated so far. On the first pass
through the scratch file, the list of energies to be used for printing information
on the listing is established in elist using a few if statements based on the
range of the energy variable e. For each energy, the corresponding cross section is
retrieved using gety1 and the appropriate E and damage numbers are computed
by calling getsix, disbar, conbar, or capdam. The heating contribution is
computed from the appropriate formula, and the heating and damage numbers
are summed into the c array at location index. If requested, the kinematic limits
on the heating are computed and summed into the c array at index+npk and
index+2*npk. The completed results for this energy and reaction are written
out onto the loada/finda scratch file, and the energy loop is continued.
When the energy loop is complete, the subroutine jumps to the next section
(or subsection in the case of File 6) and repeats the entire energy loop for that
reaction (or particle from File 6).

NJOY2016 153
LA-UR-17-20093 6 HEATR

Subroutine indx is used to select what element of the c array is to receive

the heating or damage contribution for a section with a particular MT num-
ber. The meaning of each element of the c array is obtained from the mtp
array. Normally, a reaction MT contributes to the partial heating element with
mtp(i)=MT+300. But it can also contribute to several other elements of c, such
as nonelastic (MT=303), inelastic (MT=304), etc. Therefore, indx returns the
count of reactions contributed to by mt in nmt and the indexes for the c array in
imt(nmt).
Subroutine capdam is used to compute the damage energy for neutron capture
(or disappearance) reactions; that is, for MT=102, 103, etc. On the initialization
entry (ee=0.0), the routine sets up various kinematics parameters, such as zx
and ax to describe the outgoing particle, and initializes df. In order to save
time, the routine only calculates the damage on a grid that increases by steps of
10%. Intermediate values are obtained by interpolation (see el, daml, en, and
damn). The values at the grid points are computed using
   2 !
E 1 AE
D +D +Q (223)
A+1 2MR c2 A + 1

for radiative capture (the corrections for multiple photon emission will be made
later), or using Eq. 206 with ER from Eq. 212 and a 4-point Gauss-Legendre
quadrature. The angular distribution for particle emission is taken to be isotropic.
Subroutine disbar is used by nheat to compute the average secondary energy
and damage energy for elastic scattering (MT=2), discrete-inelastic scattering
(MT=51-90), or discrete-level particle production (MT=600-648, 650-698, etc.
for ENDF/B-VI or ENDF/B-VII, or MT=700-717, 720-737, etc. for earlier ver-
sions). It starts by initializing hgtfle (which is very similar to getfle in the
GROUPR module), determining kinematic parameters like awp (the mass ratio
to the neutron for the emitted particle), and initializing df. In order to save
time, it only computes the heating and damage on a grid based on steps by a
factor of 1.1 and the enext values from hgtfle. On a normal entry, it inter-
polates between these values (see el, cl, daml, en, cn, and damn). When the
desired ee exceeds en, the old high values are moved down to the low positions,
and new high values are calculated. The calculation of cn follows Eq. 168. The
calculation of damn uses Eq. 206 with a 20-point Gauss-Legendre quadrature (see
nq, qp, and qw).
Function df is used to compute the damage partition function given in Eq. 200.
The constants that depend on the recoil atom or particle type and lattice type

154 NJOY2016
6 HEATR LA-UR-17-20093

(see zr, ar, zl, al) are computed in an initialization call with e=0.0. Thereafter,
it can be called with any other value of e.
Similarly, conbar computes the average secondary energy and damage en-
ergy for continuous distributions described in File 5. Analytic representations
use simple formulas coded into anabar or a combination of adaptive and Gaus-
sian quadrature in anadam. Tabulated data are interpolated from the File 5
table using tabbar or integrated using trapezoidal and Gaussian quadratures
in tabdam. As usual, the routine is initialized by calling it with e=0.0. The
secondary-particle yield is either chosen from the MT number, or hgtyld is ini-
tialized. The desired section of File 5 is located on the input ENDF tape, and the
kinematic constants are computed. The reactions with MT=22, 28, 32, 33, and
34 will be treated using the capdam method; if mtd has one of these values (see
mtt), capdam is initialized. As is the case for getsed in the GROUPR module,
this routine can handle some sections of File 5 that contain multiple subsec-
tions, but the analytic subsections must come first. As each analytic subsection
is read, appropriate data are stored in in the external array c using pointers
saved in the array loc. Only the first energy is read and stored for a tabulated
subsection (lf=1). The idea is to have only two energy values in memory at a
time in order to save storage; the second subsection will be read during the first
normal entry to the subroutine. The final step in the initialization pass is to
initialize df. For a normal entry into conbar, the energy-dependent fission yield
is retrieved, if needed, and the loop over subsections is entered. Each subsection
in File 5 starts with a fractional-probability record. The desired value for energy
e is computed by interpolation using the standard NJOY utility routine terpa.
For analytic subsections, the routine uses anabar to compute En , and anadam or
capdam to compute the damage energy. Note that in order to save time, anadam
is only calculated on a fairly coarse grid based on steps by a factor of 1.5. The
intermediate values are obtained by interpolation using terp1. For tabulated
subsections, ebar and dame values are normally obtained by interpolation (see
elo, flo, dlo, ehi, fhi, and dhi). However, for the first entry, or whenever
e reaches ehi, the high data are moved into the low positions, new high data
are read from the File 5 subsection, and the values for heating and damage are
computed at ehi using tabbar and either tabdam or capdam.
Subroutine hgtyld is similar to getyld in the GROUPR module. It finds
the required section on the ENDF tape and reads the entire LIST or TAB1
record into memory. On normal entries, it either computes the yield using the
polynomial formula with constants from the LIST record, or it uses terpa to

NJOY2016 155
LA-UR-17-20093 6 HEATR

interpolate for the yield in the TAB1 data.

Subroutine anabar is used to compute the average energy for a neutron de-
scribed by an analytic subsection of File 5. The simple formulas used are tab-
ulated in the ENDF format manual[9]. Similarly, anadam is used to compute
the damage energy for an analytic subsection of File 5. Only LF=9 (the Sim-
ple Maxwellian Distribution) is supported; the routine returns zero for other
laws. Note that a statement function is defined to compute the secondary en-
ergy distribution for this law (see sed). For each incident energy, the spectrum
temperature theta is retrieved using terpa, and an adaptive integration stack
is initialized with points at four secondary energies, namely, 1., .5(E−U ), θ, and
E−U , where U is a parameter that sets the maximum possible value of E ′ . The
adaptive procedure proceeds to solve Eq. 208 by subdividing this starting grid
until trapezoidal integration can be used on each panel. The inner integral over
emission cosine µ is performed using a 4-point Gauss-Legendre quadrature for
each point on the adaptive grid. The function sed is used to compute g(E ′ ), and
df is used to compute the partition function.
Subroutine tabbar is used to compute the average energy of the emitted
neutron for a tabulated subsection of File 5. It can also be used for a tabulated
subsection of File 6. This option is flagged by law negative. The trick is to
set the “stride” or “cycle” through the file to be larger than 2 (see ncyc). The
angular part of the g(E→E ′ ) table is skipped, and only the E ′ and g values are
retrieved. For File 5, this routine only works for laws 1 and 5; others cause a
fatal error message to be issued. In both of these cases, the integral over E ′
needed to compute the average energy is done analytically for each panel in the
input data using a different formula for each interpolation scheme int.
Subroutine tabdam is used to compute the damage energy for a tabulated
subsection of File 5. The integration that is needed is given in Eq. 208. The
energy grid of the tabulation is assumed to be good enough to allow trapezoidal
integration to be used for E ′ , and a 4-point Gauss-Legendre quadrature is used
for µ.
Subroutine sixbar is used to compute charged-particle average energy and
damage energy represented by using a subsection of File 6. As is common with
NJOY subroutines, sixbar is initialized by calling it with e=0.0. The initializa-
tion path is controlled by j6, which is the index to the current subsection in File
6; by irec, which is 1 when a recoil response is to be calculated, and by jrec,
which tells the routine how to get back to the next subsection after a recoil calcu-
lation. If this is not a recoil subsection, the routine jumps to statement 110 and

156 NJOY2016
6 HEATR LA-UR-17-20093

starts reading in the data for the desired subsection. If it is flagged as a recoil
(see irec), the routine backs up to the subsection describing the particle that
induced the recoil and then continues by reading in the data for that particle.
The first step is to read in the TAB1 record that contains the particle yield,
identity (zap and awp), and representation law. If this law describes a two-
body recoil distribution, the routine sets jrec for a proper return, sets irec to
back up to the corresponding direct emission subsection, and jumps back to the
beginning of the routine to do the recoil calculation.
When the code finally arrives at statement number 210, it is ready to start
processing the current subsection. It reads in the parameters for laws 3 and 6,
or the TAB2 record and the data for the first energy point for the other laws.
With the data in place, it computes the corresponding values for mean energy
and damage energy using getsix or tabsq6 and returns.
In the special case where the section contains only a single subsection that
describes a neutron, the data stored in memory will be the data for that sub-
section, and the subroutine tabbar with a negative value for the law is used to
produce the low values.
On a normal entry (e>0), sixbar checks to see whether e is in the current
interpolation range. If it is, the code jumps to statement number 400. For the
analytic laws (law=3 and law=6), it uses a direct call to getsix to compute the
mean energy and damage energy. For the tabulated laws, it interpolates for the
results using the low and high data (see elo, flo, dlo, ehi, fhi, and dhi). On
the first entry, or whenever e increases to ehi, the code moves the high data to
the low positions, and then it reads in the data for the next energy and computes
a new set of high values for mean energy and damage using getsix or tabsq6.
Subroutine getsix is used to compute the mean energy and damage energy
for one particular incident energy in a subsection of File 6. The method used
depends on the value of law and the reference frame for the subsection. The first
case in the coding is for law=1 with data in the CM system.
This case uses Eqs. 218 and 219 with an adaptive integration over E ′ . The
integration stack is contained in the arrays x and y. It is primed with x(2)=0,
and h6cm is called to compute y(2) and the next grid point epnext. The first
panel is completed by calculating y(1) and x(1)=epnext. The panel is then
divided in half, and the midpoint is tested to see if it is within tol=0.02 (i.e.,
2%) of the linearly interpolated value. If not, the midpoint is inserted in x and
y, and the new top panel [that is x(2)−x(3)] is tested. This continues until
convergence is achieved in the top panel. The contributions to the heating and

NJOY2016 157
LA-UR-17-20093 6 HEATR

damage are added into the accumulating integrals at statement number 190, and
i is decremented so that the process can be repeated for the next panel down.
When i decreases to one, the current value of epnext is used to start the next
higher E ′ panel. This loop over panels continues until the entire E ′ range has
been integrated.
The next special case is for tabulated distributions that use E, E ′ , µ ordering
in the lab system. The angular part is ignored. A simple loop over the NEP
points in g(E→E ′ ) is carried out. Trapezoidal integration is used for each panel
for both heating and damage (h and d). If nd>0, the first nd entries are discrete
energies, and the values of the integrand at those energies are added into h and
d. Finally, h and d are copied into ebar and dame.
The block of coding starting at statement number 450 is used to compute
particle mean energies for the emitted particles from two-body reactions, or to
compute the mean recoil energy for a two-body reaction (see irec>0). The
calculation follows Eq. 216. Note that the kinematic factors include awp, the
mass ratio of the emitted particle to the incident particle. The parameter beta
here is the same as R in Eq. 217. If the angular distribution in File 6 is in
Legendre form, the heating and damage integrals are performed using a 20-point
Gauss-Legendre quadrature (see nq, qp, and qw). If the angular distribution is
tabulated as f (µ) versus µ, a trapezoidal integration is used for both heating
and damage.
The final option in getsix is for lab distributions that use E, µ, E ′ ordering.
See Eq. 220. The inner integrals are computed using trapezoidal integration.
The outer integral over µ also uses trapezoidal integration on the results of the
inner integrals for each µ grid point.
Note that getsix has an irec parameter in its calling list. When this pa-
rameter is greater than zero, the angular distribution is complemented and the
charge and mass of the particle are modified to represent the recoil species. The
value of irec is controlled by sixbar.
Subroutine h6cm is used by getsix to compute the lab distribution g(E→EL′ )
of Eq. 221 using the CM data in File 6. This subroutine uses h6dis, h6ddx and
h6psp to retrieve the CM discrete, tabulated or phase-space data from the file.
These routines are basically the same as f6cm, f6dis, f6ddx and f6psp. See
GROUPR for more details.
Subroutine gheat is used to correct the heating and damage values accumu-
lated during the pass through the neutron sections. It loops through all of the
reactions in File 12 and File 13 using two ENDF-type tapes. One is the input

158 NJOY2016
6 HEATR LA-UR-17-20093

PENDF tape, which is used to retrieve cross sections for use with the photon
multiplicities in File 12. The other is a version of the input ENDF tape that
has been passed through hconvr to put the photon data in a standard form (see
Chapter 8 (GROUPR) of this manual for a more detailed discussion of conver).
This scratch tape is used to retrieve the File 12 and File 13 data. It is very com-
mon to find reaction MT=3 (nonelastic) in File 12, but this reaction has been
removed from the PENDF tape because it is redundant; that is, it is equal to
MT1−MT2. Therefore, two passes are made through the File 12 data for MT=3,
an addition pass with MT=1 from the PENDF tape, and a subtraction pass with
MT=2 from the PENDF tape. Once the desired sections on the two tapes have
been found, the subroutines gambar, capdam, and disgam are initialized.
The energy loop for gheat goes through statement number 190. For each
energy, finda is used to retrieve the partial KERMA factors as computed from
the pass through the neutron files. The yield or cross section is retrieved using
gety1 into the variable y. If necessary, the corresponding cross section x is
retrieved using gety2. For cases where an energy-dependent Q is available, it is
retrieved using terp1 on the data stored at lqx. The next two lines correct the
energy of “primary” photons (lp=2).
For radiative capture represented in File 12 (MT=102), gambar, disgam,
and/or capdam are called to return E γ and Eγ2 /(2mR c2 ) for this photon spec-
trum or discrete photon and to correct the heating and values in the c array
using Eq. 171 and the second line of Eq. 211. The capture contribution to the
total photon eV-barns is added into c(npkk-1) and the photon energy yield is
loaded into c(npkk) for each subsection. When the last subsection is reached,
the capture energy check is made using this subtotal. Note that the capture
error is loaded into c(npkk-2) for later use in calculating the kinematic limits
for photon energy production.
For other photon-production reactions, the photon eV-barns contribution is
subtracted from the energy-balance heating position, added into the total photon
energy value in c(npkk-1), and added into c(npkk) for the subtotal for a section
with multiple subsections. After all the corrections have been completed for this
energy, the revised values are written out using loada. The code then moves on
to the next reaction and repeats the entire process.
When the reaction loop has been finished, gheat checks to see if it can print
out a photon energy production check. It can do this if kinematic checks have
been requested and if MT=303 was requested in the user’s list of partial KERMA
calculations. The code reads through the loada/finda file one more time. For

NJOY2016 159
LA-UR-17-20093 6 HEATR

each energy in elist, it prints out the total photon eV-barns from c(npkk-1)
and the kinematic limits elo and ehi. If the limits are violated by more than
10%, alarms consisting of the strings ++++ or –– are printed after the eV-barns
values.
Subroutine gambar is used to compute the mean energy for continuous photon
spectra and the photon recoil correction for capture. When called with e=0.0,
it locates the desired section of File 15 on the ENDF tape and reads in the first
incident energy. On a normal entry, it checks to see if e is in the range of the
data already computed (elo, ehi, etc.), and if so, it interpolates for the desired
results. If not (or on the first real entry), it moves the high data down to the low
positions, reads in the next energy from File 15, prepares new values at the new
ehi, and checks the energy range again. The photon ebar is returned by tabbar,
and the corrections to the heating value (esqb) and damage value (esqd) from
photon production are generated using tabsqr.
Subroutine tabsqr is used to compute the average recoil energy

Eγ2
(224)
2MR c2
for radiative capture for a tabulated subsection of File 15. The corresponding
damage energy is computed at the same time. The basic secondary-energy inte-
gral is over the panels defined by the grid points given in File 15. Inside each
panel, the integral is computed using a 4-point Gauss-Legendre quadrature.
Subroutine disgam is used to compute the Eγ2 and corresponding damage
energy for a discrete capture photon. The rest-mass constant is computed by
calling disgam once with e=0.
Subroutine hout writes the new PENDF tape with the desired heating and
damage MT numbers added. It also corrects the directory in MF=1/MT=451,
and it prepares the output listing for printing. The first step is to loop through
the partial KERMA factors requested and to write the data on the loada/finda
file onto a scratch tape in ENDF File 3 format. While the first partial is being
prepared, the code matches energies in c(1) against the energy list for printing
in elist. When a match is found, the partial KERMA factors are checked
against the kinematic limits, and the variables klo or khi are set if any of the
comparisons are out of bounds. The KERMA factors, kinematic limits, and error
flags are then printed on the output listing. When all of the new sections for
File 3 have been prepared, the code updates the contents of the File 1 directory.
It then loops through the rest of the input PENDF tapes copying sections to

160 NJOY2016
6 HEATR LA-UR-17-20093

the output and inserting the new sections in the appropriate places. When the
new PENDF file has been completed, hout makes VIEWR input for a set of
plots showing the total heating and the photon production compared to their
kinematic limits in both lin-lin and log-log forms. The lin-lin plots show the
high-energy range better, and the log-log plots expand the low-energy range.

6.11 Error Messages

error in heatr***requested too many kerma mts

6 values in addition to MT=301 are allowed with kinematic checks; other-
wise, 25 can be requested. See npkmax=28. When checks are requested, the
number of words needed is 3*npk+7; otherwise, npk+3 are needed.
error in heatr***requested too many q values
Limited to 30 by the global parameter nqamax=30.
error in heatr***too much energy-dependent q data
Limited to maxqbar=10000.
error in heatr***mode conversion not allowed...
Both units must be BCD (positive) or blocked binary (negative).
error in hinit***too many mf6 reactions
A maximum of 320 reactions are allowed. See the global parameter maxmf6=320.
message from heatr–-mt301 always calculated
MT=301 was given in the input list of partial KERMA factors. This is not
necessary; it is always inserted automatically.
message from hinit–-mf4 and 6 missing, isotropy...
Cross sections were found for charged-particle levels in the 600 or 700 series
of MT numbers, but no corresponding angular distributions were found.
Isotropy is assumed to enable the calculation to proceed, but this evaluation
should be upgraded to include the proper sections of File 4 or 6.
message from hinit–-mt18 is redundant...
If MT=19 is present, MT=18 will be ignored.
message from hinit–-mt19 has no spectrum...
In some evaluations, the partial fission reactions MT=19, 20, 21, and 38
are given in File 3, but no corresponding distributions are given. In these
cases, it is assumed that MT=18 should be used for the fission neutron
distributions.
error in hinit***upper energy mismatch for ifc=... in mt=458
When using tabulated fission energy release components in mf1/mt458,
NJOY detected different values for the upper energy limit of some of the
components. This is an evaluation error.

NJOY2016 161
LA-UR-17-20093 6 HEATR

error in hinit***no tabulated fission q components found

mf1/mt458 contains no tabulated fission energy release components even
thought the LFC value was set to 1. This is an evaluation error.
error in hinit***bad LFC in mt=458
The LFC value in mf1/mt458 can only be equal to 0 or 1. This is an
evaluation error.
message from hinit–-mt458 is missing for this mat
The fission Q-value cannot be adjusted for delayed effects.
message from hinit–-photon momentum recoil used
message from hinit–-one-particle recoil approx. used
message from nheat–-changed Q from –- to –-
The fission Q-value is adjusted from the total (non-neutrino) value given in
File 3 to a prompt value using the delayed neutron energy from MF=1/MT=458.
error in nheat***binding energy for sequential n,2n needed
The user must enter special Q-values for the ENDF/B evaluation for 9 Be.
See the discussion in Section 6.7.
error in nheat***storage exceeded
Insufficient storage for diagnostic energy grid. See the global parameter
ilmax=100 at the start of the module.
error in nheat***upper energy tabulated fission q components ...
The tabulated fission energy components are tabulated up to an upper
energy value that is inconsistent with the upper energy value of the fission
cross section. This is an evaluation error.
error in conbar***nktot gt nkmax
More than 12 subsections found. See the parameter nkmax=12.
error in conbar***insufficient storage for raw endf data
The allocatable array a in nheat is too small. Increase na=10000.
error in hgtyld***illegal lnd, must be 6 or 8
The LND value in the ENDF file is not correct, only 6 or 8 are allowed.
error in hgtyld***storage exceeded
Increase nwmax in nheat. Currently 7000.
error in tabbar***coded for lf=1 and lf=5 only
Self-explanatory. Should not occur.
message from sixbar–-no distribution for mt –- ...
The ENDF-6 format allows the evaluator to describe a subsection of File
6 with “law=0”; that is, no distribution is given. Such sections are fine for
giving particle yields for gas production and similar applications, but they
are not adequate for computing heating and damage.
error in h6ddx***too many legendre terms
See nlmax=65 in h6ddx.

162 NJOY2016
6 HEATR LA-UR-17-20093

error in h6ddx***illegal lang

The allowed values for the angular law flag are 1, 2, and 11–15.
error in h6dis***illegal lang
The allowed values for the angular law flag are 1, 2, and 11–15.
error in bacha***dominant isotope not known for...
The Kalbach-systematics approach to computing angular distributions for
particle emission requires the separation energy as computed by the liquid
drop model. If the target for an evaluation is an element, it is necessary
to choose a dominant isotope that adequately represents the effect for this
element. Dominant isotopes for materials often evaluated as elements are
given in if statements in this routine. If the desired value is missing, it must
be added, and NJOY will have to be recompiled. See the corresponding
routines in GROUPR and ACER as well.
error in h6psp***3, 4, or 5 particles only
The phase-space law is defined for 3, 4, or 5 particles only.
message from hgtfle–-lab distribution changed to cm...
ENDF procedures require that two-body reactions be described in the CM
system. Some earlier evaluations claim to be in the lab system. However,
they are for relatively heavy targets, and changing to the CM frame will
cause only a small change in the results.
error in hgtfle***desired energy above highest energy...
Fault in the evaluation.
error in getco***limited to 64 legendre coefficients
The upgraded ENDF limit.
error in getco***lab to cm conversion not coded
Discrete scattering data should be in the CM system already.
message from hconvr–-mf3, mt... is missing
message from hconvr–-mf12, mt... is missing
error in hconvr–-missing mf3 mt’s, probable endf error
All these messages indicate missing sections in either mf3 or mf6. This is
an evaluation error.
message from hconvr–-gamma prod patch made for mt –-
This reflects some problems in the old ENDF-III evaluations for Cl and K,
which were also carried over to later ENDF versions.
error in hconvr***too many lo=2 gammas
See lmax=500.
error in hconvr***exceeded storage for nubar
See nnu=6000.
error in gheat***lo=2 not coded
Will not occur since lo=2 data have been transformed to lo=1 format by
hconvr.

NJOY2016 163
LA-UR-17-20093 6 HEATR

message from gheat–-no file 12 for this material

Information only.
message from gheat–-skipping mf.../mt... processed in mf6
NJOY has found photon data for a given mt in both mf6 and mf12 - mf15.
Only the mf6 data are used.
error in gambar***storage exceeded in a
Increase nd=10000 in gheat.
error in gambar***requested energy gt highest given
Probably reflects an error in the evaluation.
error in hout***nin out of order. read mfh,mth = ...
error in hout***nscr out of order. read mfh,mth = ...
These errors indicate that the various sections in the ENDF file are not
sequentially ordered. Check the ENDF file and correct it if possible.

6.12 Storage Allocation

Allocatable arrays are used for most large data blocks. Storage requirements are
dominated by the length of File 5 or File 15 for the evaluation. The loada/finda
buffer size nbuf may be decreased or increased at will. The code is currently
dimensioned as follows:
100 coarse grid points
30 auxiliary Q-values
25 partial KERMAS (7 when kinematic limits are requested)
10000 words of energy-dependent Q data
10000 maximum for File 5 or 15 raw data
7000 maximum for fission yield data
320 File 6 reactions

164 NJOY2016
7 THERMR LA-UR-17-20093

7 THERMR
The THERMR module generates pointwise neutron scattering cross sections in
the thermal energy range and adds them to an existing PENDF file. The cross
sections can then be group-averaged, plotted, or reformatted in subsequent mod-
ules. THERMR works with either the original ENDF/B-III thermal format[51]
and data files[52] (which were also used for ENDF/B-IV and -V), or the newer
ENDF-6 format[9]. Coherent elastic cross sections are generated for crystalline
materials using either parameters given in an ENDF-6 format evaluation or an
extended version of the method of HEXSCAT[32]. Incoherent elastic cross sec-
tions for non-crystalline materials such as polyethylene and ZrH can be gener-
ated either from parameters in an ENDF-6 format file or by direct evaluation
using parameters included in the THERMR coding. Inelastic cross sections and
energy-to-energy transfer matrices can be produced for a gas of free atoms, or for
bound scatterers when ENDF S(α, β) scattering functions are available. This
function has previously been performed using FLANGE-II[31]. THERMR has
the following features:

• The energy grid for coherent elastic scattering is produced adaptively so as

to represent the cross section between the sharp Bragg edges to a specified
tolerance using linear interpolation.
• The secondary energy grid for inelastic incoherent scattering when using E-
E ′ -µ ordering is produced adaptively so as to represent all structure with
linear interpolation. Discrete-angle representations are used to avoid the
limitations of Legendre expansions.
• An option to use E-µ-E ′ ordering is available. Dependences on µ and E ′
are constructed adaptively.
• Incoherent cross sections are computed by integrating the incoherent dis-
tributions for consistency.
• Free-atom incoherent scattering is normalized to the Doppler broadened
elastic scattering cross section in order to provide an approximate repre-
sentation of resonance scattering and to preserve the correct total cross
section.
• Hard-to-find parameters for the ENDF/B-III evaluations are included in
the code for the user’s convenience.
• ENDF-6 format files can be processed. This gives the evaluator more control
over the final results, because all parameters needed to compute the cross
sections are contained in the file.

This chapter describes the THERMR module in NJOY2016.0.

NJOY2016 165
LA-UR-17-20093 7 THERMR

7.1 Coherent Elastic Scattering

In crystalline solids consisting of coherent scatters — for example, graphite —
the so-called “zero-phonon term” leads to interference scattering from the var-
ious planes of atoms of the crystals making up the solid. There is no energy
loss in such scattering, and the ENDF term for the reaction is coherent elastic
scattering. The cross section may be represented as follows:

σc X
σ coh (E, E ′ , µ) = fi e−2W Ei δ(µ − µ0 ) δ(E − E ′ ) , (225)
E
Ei >E

where

Ei
µ0 = 1 − 2 , (226)
E
and the integrated cross section is given by

σc X
σ coh = fi e−2W Ei . (227)
E
Ei >E

In these equations, E is the incident neutron energy, E ′ is the secondary neutron

energy, µ is the scattering cosine in the laboratory (LAB) reference system, σc
is the characteristic coherent cross section for the material, W is the effective
Debye-Waller coefficient (which is a function of temperature), the Ei are the
so-called “Bragg edges”, and the fi are related to the crystallographic structure
factors.
It can be seen from Eq. 227 and the example in Fig. 16 that the coherent
elastic cross section is zero before the first Bragg edge, E1 (typically about 2 to 5
meV). It then jumps sharply to a value determined by f1 and the Debye-Waller
term. At higher energies, the cross section drops off as 1/E until E=E2 . It then
takes another jump and resumes its 1/E drop-off. The sizes of the steps in the
cross section gradually get smaller, and at high energies there is nothing left but
an asymptotic 1/E decrease (typically above 1 to 2 eV).
For evaluations in the ENDF-6 format, the section MF=7/MT=2 contains
the quantity Eσ coh (E) as a function of energy and temperature. The energy
dependence is given as a histogram with breaks at the Bragg energies. The cross
section is easily recovered from this representation by dividing by E. The Ei are
easily found as the tabulation points of the function, and the fi for a point can
be obtained by subtracting the value at the previous point.
For evaluations using the older ENDF/B-III thermal format, it is necessary to

166 NJOY2016
7 THERMR LA-UR-17-20093

1
10

Cross section (barns)

0
10
-3 -2 -1 0
10 10 10 10
Neutron Energy (eV)

Figure 16: Typical behavior of the coherent elastic scattering cross section for a crystalline
material as computed by THERMR. This cross section is for graphite at 293.6K.

compute the Ei and fi in THERMR. The methods used are based on HEXSCAT
and work only for the hexagonal materials graphite, Be, and BeO. The Bragg
edges are given by

h̄2 τi2
Ei = , (228)
8m
where τi is the length of the vectors of one particular “shell” of the reciprocal
lattice, and m is the neutron mass. The fi factors are given by

π 2 h̄2 X
fi = |F (τ )|2 , (229)
2mN V
shell

where the shell sum extends over all reciprocal lattice vectors of the given length,
N is the number of atoms in the unit cell, and F is the crystallographic structure
factor. The calculation works by preparing a sorted list of precomputed τi and
fi values. As τi gets large, the values of τi get more and more closely spaced. In
order to save storage and run time, a range of τ values can be lumped together
to give a single effective τi and fi . This device washes out the Bragg edges
at high energies while preserving the proper average cross section and angular
dependence. The current grouping factor is 5% (see eps in sigcoh).

NJOY2016 167
LA-UR-17-20093 7 THERMR

Lattice constants (given in sigcoh for graphite, Be, and BeO), form fac-
tor formulas (see form), Debye-Waller coefficients, and methods for computing
reciprocal lattice vectors were borrowed directly from HEXSCAT.
The energy grid for E is obtained adaptively (see coh). A panel extending
from just above one Bragg edge to just below the next higher edge is subdivided
by successive halving until linear interpolation is within a specified fractional
tolerance (tol) of the exact cross section at every point. This procedure is
repeated for every panel from the first Bragg edge to the specified maximum
energy for the thermal treatment (emax).
The code usually computes and writes out the cross section of Eq. 227, the
average over µ of Eq. 225, which is sometimes called the P0 cross section. Sub-
sequent modules can deduce the correct discrete scattering angles µ0 from the
location of the Bragg edges Ei and the factors fi from the cross section steps at
the Bragg edges (see GROUPR). Legendre cross sections can also be computed
by making a small change to the code. It is not necessary to give the P1 , P2 ,
and P3 cross sections explicitly as was done in some earlier codes or in File 4 of
the ENDF thermal tapes.

7.2 Incoherent Inelastic Scattering

In ENDF/B notation, the thermal incoherent scattering cross section is given by
r
′ σb E ′ −β/2
σ inc
(E, E , µ) = e S(α, β) , (230)
2kT E
where E is the initial neutron energy, E ′ is the energy of the scattered neutron,
µ is the scattering cosine in the laboratory system, σb is the characteristic bound
incoherent cross section for the nuclide, T is the Kelvin temperature, β is the
dimensionless energy transfer,

E′ − E
β= , (231)
kT
α is the dimensionless momentum transfer,
√
E ′ + E − 2µ EE ′
α= , (232)
AkT
k is Boltzmann’s constant, and A is the ratio of the scatter mass to the neu-
tron mass. The bound scattering cross section is usually given in terms of the
characteristic free cross section, σf ,

168 NJOY2016
7 THERMR LA-UR-17-20093

(A + 1)2
σb = σf . (233)
A2
The scattering law S(α, β) describes the binding of the scattering atom in a
material. For a free gas of scatterers with no internal structure

α2 + β 2
 
1
S(α, β) = √ exp − . (234)
4πα 4α
For binding in solids and liquids, S(α, β) for a number of important moderator
materials is available in ENDF/B File 7 format. The scattering law is given
as tables of S versus α for various values of β. Values of S for other values
of α and β can be obtained by interpolation. The scattering law is normally
symmetric in β and only has to be tabulated for positive values, but for materials
like orthohydrogen and parahydrogen of interest for cold moderators at neutron
scattering facilities, this is not true. These kinds of materials are identified by
the ENDF-6 LASYM option, and THERMR assumes that the scattering law is
given explicitly for both positive and negative values of β.
If the α or β required is outside the range of the table in File 7, the differential
scattering cross section can be computed using the short collision time (SCT)
approximation

p
E ′ /E (α − |β|)2 T
 
′ σb β + |β|
σ SCT
(E, E , µ) = p exp − − , (235)
2kT 4π α Teff /T 4α Teff 2

where Teff is the effective temperature for the SCT approximation. These tem-
peratures are available[52] for the older ENDF/B-III evaluations; they are usually
somewhat larger than the corresponding Maxwellian temperature T . For the con-
venience of the user, the values of Teff for the common moderators are included
as defaults (see input instructions). For the newer ENDF-6 format, the effective
temperatures are included in the data file. However, there is a complication.
Some evaluations give S(α, β) for a molecule or compound (in the ENDF/B-III
files, these cases are BeO and C6 H6 ). The corresponding SCT approximation
must contain terms for both atoms. The two sets of bound cross sections and
effective temperatures are included in the data statements in THERMR, and
they can be given in the new ENDF-6 format if desired.
THERMR expects the requested temperature T to be one of the temperatures
included on the ENDF/B thermal file, or within a few degrees of that value (296K
is used if 300K is requested). Intermediate temperatures should be obtained

NJOY2016 169
LA-UR-17-20093 7 THERMR

by interpolating between the resulting cross sections and not by interpolating

S(α, β).
The cross sections for incoherent inelastic scattering are computed in the
calcem subroutine. There are two possible orderings of the basic variables al-
lowed (see iform). For E-E ′ -µ ordering, the secondary energy grid for incoherent
scattering is obtained adaptively. A stack is first primed with the point at zero
and the first point above zero that can be derived from the positive and negative
values of β from the evaluator’s β grid using Eq. 231. (For free-gas scattering,
the β grid is taken to have 45 entries between 0.0 and 3500). This interval is
then subdivided by successive halving until the cross section obtained by linear
interpolation is within the specified tolerance of the correct cross section (from
sigl). The next highest energy derived from the β grid is then calculated, and
the subdivision process is repeated for this new interval. This process is contin-
ued until the β grid has been exhausted. Excess points with zero cross section
are removed before writing the spectrum into File 6. This procedure is sure to
pick up all the structure in the evaluation; giving points related to the β grid
avoids excessive work in trying to fit sharp corners introduced by breaks in the
interpolation of S(α, β). Fig. 17 shows how the procedure picks up features
resulting from the sharp excitation features in the graphite phonon frequency
distribution.
The result of this adaptive reconstruction is easily integrated by the trapezoid
rule to find the incoherent cross section at energy E.
The cross section for one particular E→E ′ is the integral over the angular
variable of Eq. 230. The angular dependence is obtained by adaptively subdi-
viding the cosine range until the actual angular function (see sig) is represented
by linear interpolation to within a specified tolerance. The integral under this
curve is used in calculating the secondary-energy dependence as described above.
Rather than providing the traditional Legendre coefficients, THERMR divides
the angular range into equally probable cosine bins and then selects the single
cosine in each bin that preserves the average cosine in the bin. These equally
probable cosines can be converted to Legendre coefficients easily when produc-
ing group constants, and they are suitable for direct use in Monte Carlo codes.
For strongly peaked functions, such as scattering for E≫kT when the result
begins to look “elastic”, all the discrete angles will be bunched together near
the scattering angle defined by ordinary kinematics. This behavior cannot be
obtained with ordinary P3 Legendre coefficients. Conversely, if such angles are
converted to Legendre form, very high orders can be used. If a direct calculation

170 NJOY2016
7 THERMR LA-UR-17-20093

30

25

20

Probability
15

10

0
0 50 100 150
Neutron Energy (eV) *10-3

Figure 17: Adaptive reconstruction of two of the emission curves for graphite at 293.6K
(E=.00016 eV to the left, and E=.1116 eV to the right). Note the presence of
excitation features from the phonon frequency spectrum for both upscatter and
downscatter. The breaks in the curves are due to β interpolation in S(α, β) and
not to the tolerances in the reconstruction process. The green curves are the
corresponding free gas results.

of Legendre components is desired, reverse the sign of nnl in calcem.

The incident energy grid is currently stored directly in the code (see egrid
in calcem). The choice of grid for σ inc (E) is not critical since the cross section
is a slowly varying function of E. However, the energy grid would seem to be
important for the emission spectra. In order to demonstrate the problem, two
perspective plots of the full energy distribution of incoherent inelastic scattering
from graphite at 293.6K are shown in Figs. 18 and 19. The second plot is simply
an expansion of the high-energy region of the distribution.
It is clear that σ inc (E, E ′ ) for one value of E ′ is a very strongly energy-
dependent function for the higher incident energies. However, as shown in
Fig. 19, the shape of the secondary energy distribution changes more slowly,
with the peak tending to follow the line E ′ =E. This behavior implies that a
relatively coarse incident energy grid might prove adequate if a suitable method
is used to interpolate between the shapes at adjacent E values. One such in-
terpolation scheme is implemented in GROUPR. The use of discrete angles is
especially suitable for this interpolation scheme.
Strictly speaking, the scattering law for free-gas scattering given in Eq. 234 is

NJOY2016 171
LA-UR-17-20093 7 THERMR

GRAPHITE AT 293.6K FROM ENDF/B-VII

thermal inelastic

7
10

1 0 -8
Prob/MeV

6
10

1 0 -9

)
5

eV
10

y (M
10 -10

e rg
1 0 -1 0
10 -9

En
Se 10 -8
c.
En
erg 10 -7

1 0 -1 1
y

Figure 18: Neutron distribution for incoherent inelastic scattering from graphite (T =
293.6K).

GRAPHITE AT 293.6K FROM ENDF/B-VII

thermal inelastic

7
10
Prob/MeV

6
10
1 0 -7
)

5
eV

10
y (M
e rg

10 -8
En

Se
c. 1
En
erg 0
-7
y

Figure 19: Expanded view of the high-energy region of the graphite incoherent inelastic
distribution.

only applicable to scatterers with no internal structure. However, many materi-

als of interest in reactor physics have strong scattering resonances in the thermal
range (for example, 240 Pu and 135 Xe). The Doppler broadened elastic cross sec-
tion produced by BROADR is formally correct for a gas of resonant scatterers,

172 NJOY2016
7 THERMR LA-UR-17-20093

but the cross section resulting from Eq. 234 is not. In order to allow for reso-
nance scattering in a way that at least provides the correct total cross section,
THERMR renormalizes the free-atom scattering to the broadened elastic cross
section. The secondary energy distribution will still be incorrect.
The built-in grid for incident neutron energies is suitable for normal temper-
atures found in reactors. For higher temperatures (higher than break=3000),
the grid values are scaled up to span the kinds of energies expected.
If the E-µ-E ′ option is selected (iform=1), an adaptive reconstruction of
the angular cross section σ(E, µ) is performed. For each µ value, the secondary
energy spectrum is generated adaptively, and the integral over that spectrum
is saved as σ(E, µ). The results are written out using the ENDF-6 format File
6/law 7 option. This ordering is more like the results of experiments, and the
THERMR results can be used to compare to experiment. See Fig. 20 for a figure
based on this kind of ordering.

20

15
Probability

10

0
0 50 100 150 200
Neutron Energy (eV) *10-3

Figure 20: Example of distributions for H in H2 O with E-µ-E ′ ordering. The incident energy
is 0.115 eV. The black curve is at 51.3 deg, the red curve is at 60 deg, and the
green curve is at 68 deg.

NJOY2016 173
LA-UR-17-20093 7 THERMR

7.3 Incoherent Elastic Scattering

In hydrogenous solids, there is an elastic (no energy loss) component of scat-
tering arising from the zero-phonon term that can be treated in the incoherent
approximation because of the large incoherent cross section of hydrogen. The
ENDF term for this process is incoherent elastic scattering, and it is found in the
materials polyethylene and zirconium hydride. The differential cross section is
given by

σb −2W E(1−µ)
σ iel (E, E ′ , µ) = e δ(E − E ′ ) , (236)
2
where σb is the characteristic bound cross section and W is the Debye-Waller
coefficient. The energy grid of the elastic cross section is used for E, and the
average cross section and equally probable angles are computed using

1 − e−4W E
 
σb
iel
σ (E) = , (237)
2 2W E
and

N h −2W E(1−µi )
µ̄i = e (2W Eµi − 1)
2W E i
− e−2W E(1−µi−1 ) (2W Eµi−1 − 1) /(1 − e−4W E ) , (238)

where

1 − e−4W E
 
1
µi = 1 + ln + e−2W E(1−µi−1 ) (239)
2W E N
is the upper limit of one equal probability bin and µ̄i is the selected discrete
cosine in this bin. Here N is the number of bins and µ0 is −1.
The characteristic bound cross sections and the Debye-Waller coefficients can
be read from MF=7/MT=2 of an ENDF-6 format evaluation, or obtained di-
rectly from data statements in the code for the older format.

7.4 Coding Details

The thermr routine comes from module thrmm. The procedure begins with the
reading of the user’s input. The required ENDF tape (nendf) is only used for
MF=7 data; it can be set to zero if only free-gas scattering is needed. Similarly,
matde is the material number for the File 7 tape and can be set to zero for

174 NJOY2016
7 THERMR LA-UR-17-20093

free-atom problems. The ENDF File 7 format only gives “M0 σf 0 ”, the product
of the free scattering cross section for the principal scatterer and the number
of principal scatterer atoms in the molecule. As a result, THERMR needs the
parameter natom to obtain the effective microscopic cross section (for example,
for H in H2 O, use natom=2). For ENDF/B-III format files, default parameters
are supplied for mixed moderators (BeO and benzine) and effective temperatures,
if needed.
Continuing, thermr finds the desired material on the input PENDF and
ENDF tapes. It will automatically loop over ntemp materials on nin. The
input tape must have been through BROADR. The elastic cross section at the
current temperature is saved on a loada/finda scratch file to be used for nor-
malizing free-atom scattering if necessary. For ENDF-6 format materials, the
parameters for the elastic calculation are read in using rdelas. Next, thermr
computes elastic and/or inelastic cross sections by calls to coh, iel, and calcem.
Finally, the results are written onto the output PENDF tape by tpend.
Some alteration of ENDF/B formats and conventions was required to accom-
modate thermal cross sections. The incoherent inelastic cross sections fit well
into MF=3 using MT=mtref (see user input). The coherent or incoherent elas-
tic cross section (if present) uses mtref+1. Other modules of NJOY expect that
thermal MT numbers will be between 221 and 250. The incoherent energy-to-
energy matrix is stored in MF=6 (coupled angle-energy distributions). Before
the introduction of the ENDF-6 format, the ENDF File 6 formats were not well-
suited to this application because secondary angle and energy were not tightly
coupled as required by the physics of the problem. Therefore, three new formats
were defined for File 6: LTT=5 for discrete-angle inelastic transfer cross sections,
LTT=6 for discrete-angle elastic data, and LTT=7 for coherent elastic reactions.
The format for LTT=5 follows in the notation of ENDF-102[51]:

MAT,6,MT [ ZA, AWR, 0, LTT, 0, 0 ] CONT LTT=5

MAT,6,MT [ T, 0., 0, 1, NNE / NNE, 2 ] TAB2
MAT,6,MT [ 0., EN, 0, 0, NEP*(NL+1), NL+1 /
EP(1), PP(1), EPM(1,1), ...
EP(2), PP(2), ... ] LIST
... repeat the LIST for the NNE values of EN ...
MAT,6,0 [ 0., 0., 0, 0, 0, 0 ] SEND

There is a list record for each of the NNE values of incident energy. Each list

NJOY2016 175
LA-UR-17-20093 7 THERMR

record gives the normalized secondary energy distribution as NEP values of PP

vs. E ′ , and for each value of E ′ , the record gives NL equally probable cosines,
EPM. Similarly, the format for LTT=6 is:

MAT,6,MT [ ZA, AWR, 0, LTT, 0, 0 ] CONT LTT=6

MAT,6,MT [ T, 0., 0, 1, NNE / NNE, 2 ] TAB2
MAT,6,MT [ 0., EN, 0, 0, NU+2, NU+2 /
EN, 1., U(1), U(2), ... ] LIST
... repeat the LIST for the NNE values of EN ...
MAT,6,0 [ 0., 0., 0, 0, 0, 0 ] SEND

Here, there is just a set of NU equally probable cosines given for each incident
energy. Note that this format was designed to look like that for LTT=5 with
NEP=1. Finally, the format for LTT=7 is:

MAT,6,MT [ ZA, AWR, 0, LTT, 0, 0 ] CONT LTT=7

MAT,6,MT [ 0,, 0., 0, 0, 0, NBRAGG ] CONT
MAT,6,0 [ 0., 0., 0, 0, 0, 0 ] SEND

In this case, all the important information is in File 3 under MT=mtref+1. For
convenience, the number of Bragg edges used is given here in File 6 as NBRAGG.
In subroutine coh, the energy grid is determined adaptively and stored into
the same loada/finda scratch file used for the elastic cross section. The elastic
cross section is converted to the coherent grid using Lagrangian interpolation
(see terp). The structure of the record stored on the scratch file is [energy /
static elastic / incoherent inelastic / coherent elastic].
Coherent cross sections at a given energy E are computed by sigcoh. If
this is the first entry (E=0) for an ENDF-III type material, the appropriate
lattice constants are selected and the Debye-Waller coefficient is obtained for the
desired temperature by interpolation. Then the reciprocal lattice wave vectors
and structure factors are computed, sorted into shells, and stored for later use.
On a normal entry (E>0), the stored list is used to compute the cross section.
For ENDF-6 format materials, the initialization step is used to organize the data
already read from MF=7/MT=2 by rdelas, and subsequent entries are used to
compute the cross section.

176 NJOY2016
7 THERMR LA-UR-17-20093

Incoherent elastic cross sections are computed in subroutine iel. The appro-
priate bound cross sections and Debye-Waller coefficients are either extracted
from the data already read from an ENDF-6 format MF=7/MT=2 by rdelas,
or they are extracted from data statements in iel and then adjusted to the
specified temperature using terp or terpa. The angle-integrated cross section is
computed analytically on the grid of the static elastic cross section and written
back onto the loada/finda scratch file in the same slot used for coherent elastic
as described above (both never occur simultaneously in the same material). The
discrete equally probable cosines are cast into LTT=7 format and written onto
a scratch tape for use by tpend.
Incoherent cross sections and distributions are generated in calcem. On the
first entry, the ENDF/B scattering law is read in or parameters are set for
free-atom scattering. For ENDF-6 files, the effective temperatures for the SCT
approximation are read in. For the older formats, these numbers were either
read in or set to default values during the user input process. The calculation
for E-E ′ -µ ordering (iform=0) goes through statement 300. An adaptive loop to
determine the secondary energy grid is carried out. The required cross sections
and discrete cosines are returned by sigl, which uses sig to compute the dif-
ferential cross sections. Because the spectrum curve will have discontinuities in
slope at energies corresponding to the break points of the β grid, it is important
to use these energies as the starting points for the adaptive reconstruction. The
first panel starts at E ′ = 0 and ends at the first energy greater than zero that
can be derived from the β grid. This will normally be a negative β value cor-
responding to E ′ <E. These two energies and their corresponding cross sections
are loaded into an inverted stack like the one used in RECONR. Next, the top
interval in the stack is divided in half, and new cross sections are computed at
this midpoint. If the new cross section is not within the desired tolerance of the
value obtained by linear interpolation between the adjacent points, the new value
is inserted into the stack. Otherwise, the top value in the stack is converged and
can be saved to the location where the spectrum is accumulating. Each time the
stack gets down to a single element, a new point is calculated from the next β
value in the evaluator’s β grid, and the subdivision process is continued. During
this reconstruction process, the integrated cross section is computed by adding
in each trapezoid. In addition, note is taken of the last nonzero cross section
value in order to remove excess zero values from the end of the record. The σ vs.
E ′ curve is complete when the β grid has been exhausted (the highest positive
value). The result is put directly into the modified MF=6 format and written

NJOY2016 177
LA-UR-17-20093 7 THERMR

onto a scratch file.

When all the desired incident energies have been processed, the incoherent
cross section is calculated on the File 3 energy grid by interpolating in the table
of values computed by the reconstruction process. The results are stored on the
loada/finda tape. If free-atom scattering has been selected, the elastic cross
section is stored in the incoherent slot.
Incoherent inelastic scattering cross sections and discrete cosines are com-
puted in sigl. The stack for the adaptive reconstruction of the angular distri-
bution for a given E→E ′ is primed with µ=−1, µ=+1, and the angle for static
scattering. The top interval on the stack is subdivided by halving until the ac-
tual cross section computed by sig is within a specified tolerance of a linear
interpolate. As each panel is converged, its area is added to the accumulating
cross section. On convergence, the fraction of the cross section corresponding to
each equally probable bin is computed, and the linearization process is repeated
to find the bin boundaries and discrete cosines. Note that Legendre coefficients
can be computed in this routine from the discrete cosines.
Subroutine sig is used to compute the actual double-differential cross section
for a given value of E, E ′ , and µ. This is done using S(α, β) (with the possible
use of the SCT approximation for large values of α or β), or using the free-gas
scattering law. Normally, s(α, β) is symmetrical in β and results are extracted
from the table at isab using |β|. However, this is not true for materials like
orthohydrogen and parahydrogen. In these cases, the LASYM parameter is
set, and explicit S(α, β) values are given for both negative and positive β. For
liquids, the presence of diffusion leads to a singularity for β = 0 and small α
(the quasi-elastic scattering peak). The normal ENDF interpolation laws do not
represent S(α, β) well in this region. Therefore, THERMR tries to determine if
the material is a liquid by looking at the small-α behavior of the β = 0 curve
(see cliq). It then extrapolates low-β curves to low-α values using a β 2 /α law.
The calculation for E-µ-E ′ ordering starts at statement number 510. In
a loop over incident energies, subroutine sigu is called to reconstruct σ(E, µ)
adaptively. The method used is the same as that described above for the energy
distribution, and the result represents the angular cross section to within a de-
sired tolerance. The inelastic cross section is obtained as the integral over the
angular cross section. Subroutine sigu obtains the secondary energy spectrum
corresponding to E and µ adaptively. It uses the beta values from the evaluation
as starting points for subdivision, just as described above for E-E ′ -µ ordering.
It uses sig to obtain the cross section for a given E, E ′ , and µ. The results of

178 NJOY2016
7 THERMR LA-UR-17-20093

this calculation are written onto nscr using the ENDF-6 File 6/Law 7 format
and passed to tpend.
Finally, tpend is called to prepare the output tape. The File 1 directory is up-
dated to account for the new sections that are being added. File 3 is located and
the cross sections stored on the loada/finda scratch file are retrieved, format-
ted, and written to the output tape. Note that the elastic cross section in MT=2
and the total cross section in MT=1 are not changed from their static values,
nor is the union grid updated. As a result, MT=221 – 250 must be considered
supplemental. Subsequent modules could ignore them or use them in place of
the static values. Also note that it is possible to run THERMR several times
with different values of mtref. The result would be one PENDF tape containing
static cross sections and cross sections for several different binding states that
can be selected at will (for example, MT=2 for static hydrogen, MT=221 for
free hydrogen, MT=222 for hydrogen in water, and MT=223 and MT=224 for
hydrogen in polyethylene, all on one PENDF tape).
File 6 distributions are read from a scratch file (nscr) in ENDF format, nor-
malized, and written back onto the final tape. Since free incoherent scattering
was set equal to elastic scattering in calcem, the approximate resonance correc-
tion of the matrix is now complete.

7.5 Using the ENDF/B Thermal Data Files

The thermal data files originally prepared for ENDF/B-III were also used for
ENDF/B-IV and ENDF/B-V.4 Table 3 summarizes the contents of these data
tapes.
For most of these evaluations, THERMR will produce cross sections appro-
priate for the major scattering atom bound in a particular material, for example,
hydrogen bound in water, or Zr bound in ZrH. In these cases, the cross sections
are combined later (for example, hydrogen bound in water would be combined
with free-gas oxygen, and Zr bound in ZrH would be combined with H bound
in ZrH). The treatment to be used for the secondary scattering atoms for each
evaluation is indicated in the table.
In two cases – BeO and Benzine – the scattering laws S(α, β) for the two
component atoms have been combined into a single scattering law normalized to
be used with the cross section of the primary scattering atom. For these cases,
THERMR produces a cross section for the molecule or compound directly. In
4
The data files and the reference manual[52] are available from the National Nuclear Data Center,
Brookhaven National Laboratory, Upton, NY 11973. Request Tapes 320 through 325 and report ENDF-269.

NJOY2016 179
LA-UR-17-20093 7 THERMR

Table 3: Moderator Materials on ENDF/B-III Thermal Data Tapes Showing Their MAT
Numbers and Distribution Tape Numbers

Material MAT Tape MTs elastic secondary

Be 1064 321 231,232 coh
BeO 1099 321 233,234 coh none
C (graphite) 1065 322 229,230 coh
H(CH2 ) 1114 322 223 iel free C
C6 H6 1095 325 227 none
D(D2 O) 1004 320 228 free O
H(H2 O) 1002 320 222 free O
Zr(ZrHn ) 1096 323 235,236 iel 1097
H(ZrHn ) 1097 323 225,226 iel 1096

making a macroscopic cross section for BeO, the user would multiply the thermal
BeO cross section from THERMR by the atomic density of Be, taking care not
to add any additional thermal contribution for the oxygen.

Table 4: ENDF/B-VII Moderator Materials with Their MAT and MT Numbers

Material MAT MTs elastic secondary

Al 45 243,244 coh
Be 26 231,232 coh
Be(BeO) 27 233,234 coh
O(BeO) 28 237,238 coh
C(graphite) 31 229,230 coh
Fe 56 245,246 coh
H(CH2 ) 37 223 iel free C
H(liquidCH4 ) 33
H(solidCH4 ) 34
C6 H6 40 227 none
D(D2 O) 11 228 free O
D(paraD2 ) 12
D(orthoD2 ) 13
H(H2 O) 1 222
H(paraH2 ) 2
H(orthoH2 ) 3
Zr(ZrHn ) 58 235,236 iel
H(ZrHn ) 7 225,226 iel
U(UO2 ) 76 241,242 coh
O(UO2 ) 75 239,240 coh

180 NJOY2016
7 THERMR LA-UR-17-20093

THERMR labels the thermal cross sections that it generates with specially
defined MT numbers. The particular numbers shown in the table are recognized
by the reaction naming logic in MATXSR. Note that two numbers are defined
for materials that have both inelastic and elastic components; the first number
is for inelastic, and the second for elastic.
For ENDF/B-VII, there are additional thermal materials available, and the
numbering has changed. See Table 4. Other thermal moderators are expected
in future ENDF/B generations as well as in other regional libraries (e.g., JEFF,
JENDL, etc.).

7.6 Input Instructions

The following input instructions have been copied from the comment cards in
THERMR:

!---input specifications (free format)-------------------

!
! card 1
! nendf endf tape for mf7 data
! nin old pendf tape
! nout new pendf tape
! card 2
! matde material desired on endf tape
! matdp material desired on pendf tape
! nbin number of equi-probable angles
! ntemp number of temperatures (default=1)
! iin inelastic options
! 0 none
! 1 compute as free gas
! 2 read s(a,b) and compute matrix
! icoh elastic options
! 0 none
! 1 compute using ENDF6 format data
! --------or for earlier formats
! 1 graphite
! 2 beryllium
! 3 beryllium oxide
! 11 polyethylene
! 12 h(zrh)
! 13 zr(zrh)

NJOY2016 181
LA-UR-17-20093 7 THERMR

! iform output format for inelastic distributions

! 0 E-E’-mu ordering (MF6 special)
! 1 E-mu-E’ ordering (MF6/Law7)
! natom number of principal atoms
! mtref mt for inelastic reaction (221-250 only)
! iprint print option (0=minimum, 1=maximum,
! 2=max. normal + intermediate results)
! (default=0)
! card 3
! tempr temperatures (kelvin)
! card 4
! tol tolerance
! emax maximum energy for thermal treatment
! (for temperatures greater than 3000,
! emax and the energy grid are scaled by
! temp/3000. free gas only.)
!
! nendf can be endf-6 format (e.g., from leapr) while
! nin and nout are endf-4 or 5 format, if desired.
!
!-----------------------------------------------------------

The following sample problem illustrates the production of thermal cross

sections for hydrogen in water.

thermr
20 21 22/
1 125 8 2 2 0 0 2 222 0/
293.6 500/
.01 4.6/
stop

It is assumed that ENDF/B-VII evaluation for H in H2 O is mounted on unit

20, and that a previously prepared PENDF file for 1 H (MAT125) is mounted
on unit 21. The thermal cross sections for hydrogen in water will be written on
unit 22 using mtref=222. Note that the parameter natom is set to 2 because
the water molecule H2 O contains two hydrogen atoms. In addition, icoh=0 for
hydrogen in water, and iform=0 for E-E ′ -µ ordering. The higher-energy parts

182 NJOY2016
7 THERMR LA-UR-17-20093

of the neutron emission curves for this example are shown in Fig. 21. The sharp
peak at E=E ′ is quasi-elastic scattering broadened by diffusion.

1-H-1 IN H2O AT 293.6K FROM ENDF/B-VII USING CONTINUOUS

thermal inelastic

7
10
Prob/MeV

6
10

1 0 -6
)
5

eV
10

y (M
e rg
En
10 -7
Se
c.
En
erg 10 -6
y

Figure 21: Expanded view of the high-energy region for the incoherent inelastic distribution
of hydrogen bound in water. Note the sharp quasi-elastic peak at E=E ′ .

A calculation of both free and graphite cross sections for ENDF/B-VII carbon
would go as follows:

reconr
20 22 / tape20 is carbon
/
600 1/
.001/
’6-C-nat from ENDF/B-VII’/
0/
broadr
20 22 23/
600 1/
.001/
293.6/
0/
thermr
0 23 24/

NJOY2016 183
LA-UR-17-20093 7 THERMR

0 600 8 1 2 0 0 1 221 0/
293.6/
.005 5/
thermr
26 24 25 / tape 26 is ENDF/B-VII graphite
31 600 8 1 2 1 0 1 229 0/
293.6/
.005 5/
stop

First, ENDF/B-VII carbon must be mounted on unit 20. At LANL this is done
by copying it to a file named tape20 in the user’s local file space. Similarly,
ENDF/B-VII graphite must be copied to tape26. Next, RECONR is run to
linearize the evaluation, and BROADR is run to prepare 293.6K cross sections.
The first THERMR run is for free-gas scattering (mtref=221), and the second
run is for carbon bound in graphite (mtref=229). Note that natom is now 1, and
that 8 discrete angles were requested in both cases. For graphite, icoh is set to 1
in order to request the calculation of coherent elastic scattering like that shown
in Fig. 16. The coherent results will use MF=3 and MT=230. Distributions will
be calculated to 0.5% accuracy for energies up to 5 eV. The final PENDF file
will be tape25.

7.7 Error Messages

error in thermr***nin=0
An input PENDF tape is required.
error in thermr***mode conversion not allowed
nin and nout must both be binary or both be coded.
error in thermr***illegal reference mt
Restricted to MT=221–250. User’s should note that Card 2 has an ad-
ditional input parameter, iform, beginning with NJOY2012. Unmodified
NJOY99 input decks will produce this error message.
error in thermr***desired material not on pendf tape
Check input instructions against contents of thermal tape.
error in thermr***desired temperature not on tape
Check input instructions against contents of thermal tape.
error in rdelas***too much elastic data
There is not enough space in the allocatable array a. See na=10000.

184 NJOY2016
7 THERMR LA-UR-17-20093

error in rdelas***desired temp not found

The temperatures requested in THERMR must be the same as the temper-
atures on the input PENDF tape.
error in coh***too many legendre orders
The code currently computes only P0 , but nl=1 in coh can be changed if
desired. Code is currently limited to 6 (P5 ). If more coefficients are desired,
increase nlmax and the dimensions of the variables s, ej, and ex in coh,
calcem, and tpend.
error in sigcoh***storage exceeded
Not enough room for lattice factors. Increase nw which is currently 10000.
error in sigcoh***illegal lat
Only three lattices are coded so far. To add others, insert the constants in
sigcoh and form factor formulas in form.
error in iel***bad temperature for debye-waller factor
Shouldn’t occur unless new materials are added incorrectly.
error in iel***unknown material identifier
Only three options are coded so far. To add others, insert data statements
for the Debye-Waller integrals and values for the bound cross sections.
error in calcem***nl too large for binning
Increase the value of the parameter nlmax which is currently 33.
error in calcem***storage exceeded
There is not enough space in the global scratch array. Increase nwscr, which
is currently 99000, in thermr.
error in calcem***desired temperature not found
Any temperature requested in THERMR must be on the input PENDF
tape.
error in calcem***bad temperature for teff
error in calcem***bad temperature for teff2
Data for effective temperatures use a temperature grid that is not consistent
with the S(α, β) data.
error in calcem***isabt=1 pendf tape found
THERMR cannot process this format.
error in calcem***only 2 sct atoms allowed
For mixed moderators, such as BeO and Benzine, the SCT contribution
from each atom must be included. The code only allows for 2.
error in calcem***too many angles
For E-µ-E ′ ordering, there are more than mumax=300 angles being gener-
ated.
error in sig***illegal option
Only tabulated S(α, β) and free gas are coded at this time.

NJOY2016 185
LA-UR-17-20093 7 THERMR

error in sigl***no legal solution

message from sigl–-disc=-ffff, set to abs value....
error in sigl***no legal solution (quadratic path)
The code has trouble solving the equation for the boundary of a bin.
error in tpend***storage exceeded
Increase nwscr in thermr.
error in tpend***cross section = 0
Thermal cross section of zero cannot be used to normalize the distribution.

7.8 Input/Output Units

The following logical units are used.

10/11 iold/inew in thermr. Also used in coh, calcem, and tpend. Used for
the loada/finda scratch file that saves the energy grid and reaction
cross sections.
12 nscr in thermr. Also used in calcem and tpend. Contains the scat-
tering matrix before normalization.
13 nscr2 in thermr and tpend. Contains data from nin that are to be
simply copied to nout.
20-99 User’s choice for nendf, nin, nout, and nread (iinc=2 only) to link
with other modules. No mode conversion between nin and nout
allowed.

Units 10 and 11 are always binary. Units 12 and 13 have the same mode as
nin and nout. The user can choose the modes for nendf, nin, nout, except nin
and nout must have the same mode.

7.9 Storage Allocation

The storage allocated in THERMR is for the loada/finda buffers and a scratch
array. The value of nbuf my be changed at will; larger values increase I/O
efficiency. The variable nwscr controls the maximum size of the TAB1 records
of σ(E → E ′ ) versus E ′ for incoherent scattering. Hence it interacts with tol.
The linearization stack (stk) in coh is controlled by imax and the number of
Legendre components requested (always 1 in the standard version). The current
value of imax (20) is sufficient to divide each panel into parts as small as one-
millionth of the panel size. The length of the list of lattice factors (fl) in sigcoh
is controlled by the size of the ENDF/B File 7, and nw=10000 must be big enough
for the problem.

186 NJOY2016
8 GROUPR LA-UR-17-20093

8 GROUPR
GROUPR computes group-to-group scattering matrices, and anisotropic photon
production matrices for neutrons from ENDF/B-IV and later evaluated nuclear
data. With ENDF-6 format files, photonuclear data and incoming and outgoing
charged particles can also be handled. Special features are provided for ratio
quantities (for example, µ, ν, or photon yield), inverse velocity, delayed neutron
spectra by time group, and anisotropic thermal neutron scattering. Fission is
represented as a group-to-group matrix for full generality. Scattering matrices
and photon production matrices may be self-shielded if desired.
The Bondarenko narrow-resonance weighting scheme[39] is normally used.
Optionally, a weighting flux can be computed for various mixtures of heavy
absorbers with light moderators. An accurate pointwise solution of the integral
slowing down equation is used. This option is normally called on to account for
intermediate resonance effects in the epithermal range.
Neutron data and photon-production data are processed in a parallel manner
using the same weight function and quadrature scheme. This assures consistent
cross sections for coupled neutron-photon problems. Two-body scattering is
computed with a center-of-mass (CM) Gaussian quadrature, which gives accurate
results even for small Legendre components of the group-to-group matrix.
User conveniences include free-form input and complete control over which
reactions are processed. The neutron group structure, photon group structure,
and weight function can each be read in or set to one of the internal options. Out-
put can be printed and/or written to an output “groupwise-ENDF” (GENDF)
file for further processing by a formatting module (DTFR, CCCCR, MATXSR,
WIMSR), by the covariance module (ERRORR), or by the MCNP continuous-
energy Monte Carlo module (ACER).
This chapter describes the GROUPR module in NJOY2016.0.

8.1 Multigroup Constants

Multigroup constants are normally used by computer codes that calculate the
distributions of neutrons and/or photons in space and energy, and that compute
various responses to these distributions, such as criticality, dose to personnel, or
activation of materials. These distributions are solutions of the neutral particle
transport equation.5
5
The following development uses a notation based on Bell and Glasstone[53], where the lower-case sigma
is used for both macroscopic and microscopic cross sections, depending on the context. One-dimensional

NJOY2016 187
LA-UR-17-20093 8 GROUPR

∂
µ ϕ(x, µ, E) + σt (x, E) ϕ(x, µ, E)
∂x Z Z
′
= dΩ dE ′ σX (x, E ′ →E, Ω′ →Ω) ϕ(x, µ′ , E ′ )
+ Q(x, µ, E) , (240)

where the flux ϕ is allowed to vary with position x, direction Ω with polar cosine
µ, and energy E. Similarly, the macroscopic total cross section σt varies with
position and energy. The right-hand side of the equation contains the source
due to transfers from other directions Ω′ and energies E ′ (as described by the
macroscopic transfer cross section σX ), and a fixed or external source Q.
The macroscopic cross sections (in units of cm−1 ) in Eq. 240 can be calculated
from microscopic cross sections for the component isotopes or elements (in barns)
using

X
σt (x, E) = ρi (x) σti (T [x], E) , (241)
i

where ρi is the number density for a constituent (in barns−1 cm−1 ), which may
vary with position, and T is the temperature, which may also vary with position.
A similar formula holds for σX .
The transfer cross section σX (which includes both scattering and fission
processes) is normally assumed to depend only on the cosine of the scattering
angle, µ0 = Ω · Ω′ . This allows σX to be expanded using Legendre polynomials

∞
X 2ℓ + 1
σX (x, E ′ →E, Ω′ →Ω) = σXℓ (x, E ′ →E) Pℓ (µ0 ) . (242)
4π
ℓ=0

Application of the addition theorem and integration over azimuthal angle then
gives

∂
µ ϕ(x, µ, E) + σt (x, E) ϕ(x, µ, E)
∂x
∞ Z
X 2ℓ + 1
= Pℓ (µ) σXℓ (x, E ′ →E) ϕℓ (x, E ′ ) dE ′
2
ℓ=0
+ Q(x, µ, E) , (243)
slab geometry is used throughout for simplicity.

188 NJOY2016
8 GROUPR LA-UR-17-20093

where
Z
ϕℓ (x, E) = Pℓ (µ) ϕ(x, µ, E) dµ . (244)

The desired responses are then given by

Z
R(x) = σr (x, E) ϕ0 (x, E) dE , (245)

where σr is the reaction cross section for the response. The next step is to
integrate Eqs. 243 and 245 over a range of energies chosen to lie in group g. The
results are

∞
∂ X
µ ϕg (x, µ) + Pℓ (µ) σtℓg (x) ϕℓg (x)
∂x
ℓ=0
∞
X 2ℓ + 1 X
= Pℓ (µ) σXℓg′ →g (x) ϕℓg′ (x) + Qg (x, µ) ,(246)
2
ℓ=0 g′

and

X
R(x) = σrg (x) ϕ0g (x) , (247)
g

where
Z
ϕℓg (x) = ϕℓ (x, E) dE , (248)
g
Z
σt (x, E) ϕℓ (x, E) dE
g
σtℓg (x) = Z , (249)
ϕℓ (x, E) dE
g
Z
σr (x, E) ϕ0 (x, E) dE
g
σr (x) = Z , (250)
ϕ0 (x, E) dE
g

and Z Z
dE dE ′ σXℓ (x, E ′ →E) ϕℓ (x, E)
g g′
σXℓg′ →g = Z . (251)
ϕℓ (x, E) dE
g

NJOY2016 189
LA-UR-17-20093 8 GROUPR

The last three equations provide the fundamental definitions for the multi-
group cross sections and the group-to-group matrix. Note that the values of
the group constants depend upon the basic energy-dependent cross sections ob-
tained from an ENDF-format evaluation by way of the RECONR, BROADR,
UNRESR, HEATR, THERMR, and PURR modules of NJOY, and the shape of
ϕ within the group.

8.2 Group Ordering

Since neutrons normally lose energy in scattering, the scattering source into
group g depends on the flux at higher-energy groups g ′ and the cross section for
transferring neutrons from g ′ to g. For this reason, Eq. 246 is usually solved by
sweeping from high energies to low energies. (Any thermal upscatter or fission
is handled by iteration.) Data libraries for use with transport codes normally
number the groups such that group 1 is the highest-energy group, and all the
scattering matrix elements that transfer neutrons into group 1 are given first,
followed by those for scattering into group 2, and so on.
However, in ENDF files the evaluated nuclear data are always given in order
of increasing incident energy, and secondary neutron distributions are described
by giving emission spectra for given incident energies. Therefore, GROUPR
numbers its groups such that group 1 is the lowest-energy group, and it calculates
that scattering out of group 1, followed by the scattering out of group 2, and so
on.
The “backward” energy-group numbering convention used by GROUPR is a
possible source of confusion in interpreting output produced by the various mod-
ules of NJOY. All group indices printed by GROUPR or written to the GROUPR
output file use the increasing-energy order. The covariance modules and photon
interaction module follow the GROUPR ordering convention. Output modules
such as DTFR, CCCCR, MATXSR, and WIMSR invert the group order and
rearrange the scattering matrices from the GROUPR outscatter organization to
the “transport” inscatter form. Any group indices printed by these three output
modules will be in the conventional transport decreasing-energy order.

8.3 Basic ENDF Cross Sections

The basic energy-dependent cross sections and energy-angle distributions needed
for Eqs. 249, 250, and 251 are obtained from evaluated nuclear data in ENDF
format[9]. These data are indexed by material (MAT), type of information (MF),

190 NJOY2016
8 GROUPR LA-UR-17-20093

and reaction (MT). Materials can be single isotopes, elements, or compounds.

Type of information includes energy-dependent cross section (MF=3), angular
distributions (MF=4), secondary energy distributions (MF=5), and energy-angle
distributions (MF=6). Reactions include the total (MT=1) required for Eq. 249,
the partial scattering reactions that must be summed for Eq. 251 [that is, elas-
tic (MT=2), discrete-level inelastic (MT=51 – 90), (n,2n) (MT=16), etc.], and
the many partial reactions that can be used to calculate responses [for exam-
ple, (n,2n) or (n,α) for activation, gas production, heat production (KERMA),
radiation damage (DPA), etc.].
Before using GROUPR, the basic ENDF/B cross sections should have been
converted into energy- and temperature-dependent pointwise cross sections in
PENDF (pointwise ENDF) format using RECONR and BROADR. For heavy
isotopes, unresolved self-shielding data should have been added to the PENDF
tape6 using UNRESR or PURR. If requested, heat production cross sections
(KERMA), radiation damage production (or DPA), and thermal upscatter data
can also have been added to the PENDF tape using HEATR and THERMR.
See the chapters on these other modules for more information.
The detailed methods used for evaluating Eqs. 249, 250, and 251 from ENDF
and PENDF tapes are given below (see Sections 8.11 through 8.15). An example
of an ENDF pointwise cross section compared with group-averaged cross sections
from GROUPR is given in Fig. 22.

8.4 Weighting Flux

In general, the weighting flux ϕ is not known; it is, after all, the particle distri-
bution being sought in the transport calculation. However, it is often possible to
obtain fairly accurate group constants for a particular application if the shape
of the flux is reasonably well known over the broad energy ranges of a particular
few-group structure (for example, a fission spectrum, thermal Maxwellian, or
1/E slowing-down spectrum). Alternatively, one can use many small groups so
that mistakes in guessing the shape inside the group are not very important.
The key to using the multigroup method effectively is balancing the tradeoffs
between the choice of weight function and the number of groups used for each
different class of problem being solved.
In many cases of practical interest, the flux ϕ will contain dips correspond-
ing to the absorption resonances of the various materials. In the reaction rate
6
The term “tape” is used loosely in this report to refer to any input or output file. Of course, such files
would be on disk storage in a modern system.

NJOY2016 191
LA-UR-17-20093 8 GROUPR

Figure 22: A comparison of the pointwise and multigroup representations for the total cross
section of 7 Li. The Los Alamos 30-group structure is shown.

σ(E) × ϕ(E), these dips clearly reduce (self-shield) the effect of the correspond-
ing resonance. GROUPR provides two methods to estimate the effect of this
self-shielding: the Bondarenko model and the flux calculator.
In the Bondarenko model[39], the narrow resonance (NR) approximation, and
the BN approximation for large systems[53] are invoked to obtain

Wℓ (E)
ϕℓ (E) = , (252)
[σt (E)]ℓ+1
where ϕℓ is the ℓ-th Legendre component of the angular flux, the Wℓ (E) are
smooth functions of energy (such as 1/E+fission), and σt (E) is the total macro-
scopic cross section for the material. GROUPR takes all of the Wℓ to be equal to
the single function C(E), where C(E) can be read in or set to one of several inter-
nally defined functions. It is further assumed that the important self-shielding
effect of the flux can be obtained for isotope i by representing all the other
isotopes with a constant “background cross section”, σ0 . Therefore,

C(E)
ϕiℓ (E) = , (253)
[σti (E)
+ σ0i ]ℓ+1
where σti is the microscopic total cross section for isotope i. The qualitative
behavior of Eq. 253 is easy to understand. If σ0 is larger than the tallest peaks
in σt , the weighting flux ϕ is approximately proportional to the smooth weighting
function C(E). This is called infinite dilution; the cross section in the material

192 NJOY2016
8 GROUPR LA-UR-17-20093

of interest has little or no effect on the flux. On the other hand, if σ0 is small
with respect to σt , the weighting flux will have large dips at the locations of the
peaks in σt , and a large self-shielding effect will be expected.
Each component material of a mixture has a different weight function. The
macroscopic total cross sections are given by

X
σtℓg = i
ρi σtℓg (σ0i , T ) , (254)
i

where

σti (E, T )
Z
i ℓ+1
C(E) dE
g [σ0 + σt ]
i
σtℓg (σ0 , T ) = Z . (255)
1
i ℓ+1
C(E) dE
g [σ0 + σt ]

Similar equations are used for σR and σXℓg→g′ . On the GROUPR level, σ0 and T
are simply parameters. Subsequent codes, such as 1DX[54] or TRANSX[12, 41],
can compute an appropriate value for σ0 , and then interpolate in tables of cross
section versus σ0 and T to get the desired self-shielded group constants.
The appropriate value for σ0i is obvious when a single resonance material is
mixed with a moderator material (for example, 238 UO
2 ), because the admixed
materials typically have a constant cross section in the energy range where the
heavy isotopes have resonances. For a mixture of resonance materials, the normal
procedure is to preserve the average of Eq. 253 in each group by using

1 X j j
i
σ0g = ρj σt0g (σ0g ,T) , (256)
ρi
j̸=i

where the ρi are atomic densities or atomic fractions. Eq. 256 is solved by
iteration. Interference between resonances in different materials is handled in an
average sense only.
In the unresolved energy range (see UNRESR and PURR), the explicit de-
pendence of cross section on energy is not known. The integrands are replaced
by their expected values
Z  
σx
ℓ+1
Wℓ dE
g [σ0 + σt ]
σxℓg (σ0 ) = Z   , (257)
1
ℓ+1
W ℓ dE
g [σ0 + σt ]

where the expected values are averages over the distributions of resonance posi-

NJOY2016 193
LA-UR-17-20093 8 GROUPR

tion and width expected in the vicinity of energy E. The UNRESR and PURR
modules produce effective self-shielded point cross sections defined by
 
σx
[σ0 + σt ]ℓ+1
(258)


σx ℓ =   .
1
[σ0 + σt ]ℓ+1
Substituting Eq. 258 into Eq. 257 gives an equation of the form of Eqs. 249 and
250, except that σ is replaced by σ , and the flux is replaced by an average



effective flux. This effective flux can be obtained by manipulating the effective
total cross section as follows:
   
σ0 + σt − σ0 1
[σ0 + σt ]ℓ+1 [σ + σt ]ℓ
 = 0 (259)


σt ℓ =   − σ0 ,
1 1
[σ0 + σt ]ℓ+1 [σ0 + σt ]ℓ+1
from which
 
1
 
1 [σ0 + σt ]ℓ
= . (260)
[σ0 + σt ]ℓ+1


σ0 + σt k
Eq. 260 defines a recursion relation which can be used to compute the effective
flux to any order

  ℓ
1 Y 1
=
 , (261)
[σ0 + σt ]ℓ+1 σ 0 + σt k
k=0

This equation reduces to Eq. 253 in the resolved range. It is the formula used in
genflx to compute ϕℓ (E) for the Bondarenko option.
When heterogeneity effects are important, the background cross section method
can be extended as follows. In an infinite system of two regions (fuel and mod-
erator), the neutron balance equations are

Vf σf ϕf = (1 − Pf )Vf Sf + Pm Vm Sm , (262)

and

Vm σm ϕm = Pf Vf Sf + (1 − Pm )Vm Sm , (263)

where Vf and Vm are the region volumes, σf and σm are the corresponding total

194 NJOY2016
8 GROUPR LA-UR-17-20093

macroscopic cross sections, Sf and Sm are the sources per unit volume in each
region, Pf is the probability that a neutron born in the fuel will suffer its next
collision in the moderator, and Pm is the probability that a neutron born in the
moderator will suffer its next collision in the fuel. As usual, use is made of the
reciprocity theorem,

Vf σf Pf = Vm σm Pm , (264)

and the Wigner rational approximation to the fuel escape probability,

σe
Pf = , (265)
σe + σf
where σe is a slowly varying function of energy called the escape cross section,
to obtain an equation for the fuel flux in the form

σe Sm
(σf + σe )ϕf = + Sf . (266)
σm
In the limit where the resonances are narrow with respect to both fuel and
moderator scattering, the source terms Sf and Sm take on their asymptotic
forms of σp /E and σm /E respectively, and this equation becomes equivalent to
the Bondarenko model quoted above with

σe
σ0f = , (267)
ρf
and

σe + σp
C(E) = . (268)
ρf E
Note that a large escape cross section (a sample that is small relative to the
average distance to collision), corresponds to infinite dilution as discussed above.
To illustrate the general case, consider a neutron traveling through a lump of
uranium oxide with an energy close to a resonance energy. If the neutron scatters
from an oxygen nucleus, it will lose enough energy so that it can no longer react
with the uranium resonance. Similarly, if the neutron escapes from the lump,
it can no longer react with the uranium resonance. The processes of moderator
scattering and escape are equivalent in some way. Comparing Eq. 267 with
Eq. 256 gives an “equivalence principle” that says that a lump of particular
dimensions and a mixture of particular composition will have the same self-
shielded cross sections when the narrow resonance approximation is valid. The

NJOY2016 195
LA-UR-17-20093 8 GROUPR

104

Infinite
10000 barns
103 1000 barns
Cross section (barns) 100 barns
10 barns
Pointwise
2
10

101

100

10-1
101
Energy (eV)

Figure 23: The self-shielding effect on the first three 238 U capture resonances at room tem-
perature in the 5 to 50 eV range. The multigroup boundaries are from the Los
Alamos 187-group structure.

effects of material mixing and escape can simply be added to obtain the effective
σ0 for a lump containing admixed moderator material. Therefore, Eq. 256 is
extended to read
 
1 X j j
i
σ0g = σe + ρj σt0g (σ0g ,T) , (269)
ρi
j̸=i

where the escape cross section for simple convex objects (such as plates, spheres,
or cylinders) is given by (4V /S)−1 , where V and S are the volume and surface
area of the object, respectively. Many codes that use the background cross
section method modify the escape cross section as defined above to correct for
errors in the Wigner rational approximation (“Bell factor”, “Levine factor” ), or
to correct for the interaction between different lumps in the moderating region
(“Dancoff corrections”). These enhancements will not be discussed here.
As an example of self-shielded cross sections and how they vary with the
background cross section, Fig. 23 shows the first three capture resonances of 238 U
(which are very important for thermal power reactors) at room temperature for
σ0 values ranging from infinity down to 10 barns. Background cross sections
that range between 20 and 50 barns are typical for uranium-oxide pin cells.
The BROADR chapter of this report showed its capability to compute stan-

196 NJOY2016
8 GROUPR LA-UR-17-20093

dard resonance integrals. However, when self shielding and material temperature
come into play, the effective resonance integral changes. GROUPR can calculate
these quantities by doing one-group calculations with 1/E weighting over a stan-
dard energy range. Using the range 0.5 eV to the upper limit of the evaluation
will match what BROADR does. Fig. 24 shows the temperature dependence of
the capture resonance integral for 238 U from ENDF/B-VII at several different
background cross sections. The range between 20 and 50 barns is typical for
reactor pin cells.

8.5 Flux Calculator

This narrow-resonance approach is quite useful for practical fast reactor prob-
lems. However, for nuclear systems sensitive to energies from 1 to 500 eV,
there are many broad- and intermediate-width resonances which cannot be self-
shielded with sufficient accuracy using the Bondarenko approach. The GROUPR
flux calculator is designed for just such problems.
Consider an infinite homogeneous mixture of two materials and assume isotropic
scattering in the center-of-mass system. The integral slowing-down equation be-
comes

1000 barns
100 barns
50 barns
Capture Resonance Integral

20 barns
1 barn

102

101

200 400 600 800 1000 1200 1400 1600 1800

Temperature (K)

Figure 24: The capture resonance integral for 238 U showing its variation with temperature
and background cross section σ0 . The resonance integral at infinite dilution is
275.58 barns. Note the slope with temperature, which helps to produce a negative
temperature coefficient for uranium systems.

NJOY2016 197
LA-UR-17-20093 8 GROUPR

E/α1
σs1 (E ′ )
Z
σ(E) ϕ(E) = ϕ(E ′ ) dE ′
E (1 − α1 )E ′
E/α2
σs2 (E ′ )
Z
+ ϕ(E ′ ) dE ′ . (270)
E (1 − α2 )E ′

Furthermore, assume that material 1 is a pure scatterer with constant cross

section and transform to the σ0 representation. The integral equation becomes

Z E/α1
σ0
[σ0 + σt2 (E)] ϕ(E) = ϕ(E ′ ) dE ′
E (1 − α1 )E ′
E/α2
σs2 (E ′ )
Z
+ ϕ(E ′ ) dE ′ . (271)
E (1 − α2 )E ′

Finally, assume that the moderator (material 1) is light enough so that all the
resonances of material 2 are narrow with respect to scattering from material 1.
This allows the first integral to be approximated by its asymptotic form, 1/E.
More generally, the integral is assumed to be a smooth function of E given by
C(E). In this way, material 1 can represent a mixture of other materials just
as in the Bondarenko method. Fission source and thermal upscatter effects can
also be lumped in C(E). The integral equation has now been reduced to

E/α
σs (E ′ )
Z
[σ0 + σt (E)] ϕ(E) = C(E) σ0 + ϕ(E ′ ) dE ′ . (272)
E (1 − α)E ′

This is the simplest problem that can be solved using the flux calculator. The
results still depend on the single parameter σ0 , and they can be used easily by
codes that accept Bondarenko cross sections.
For heterogeneous problems, when the narrow-resonance approximation fails,
both Sf and Sm in Eq. 266 will show resonance features. To proceed further
with the solution of this equation, it is necessary to eliminate the moderator
flux that is implicit in Sm . As a sample case, consider a fuel pin immersed in a
large region of water. The fission neutrons appear at high energies, escape from
the pin, slow down in the moderator (giving a 1/E flux), and are absorbed by
the resonances in the pin. In this limit, any dips in the moderator flux caused
by resonances in the fuel are small. On the other hand, in a closely packed
lattice, the flux in the moderator is very similar to the flux in the fuel, and

198 NJOY2016
8 GROUPR LA-UR-17-20093

resonance dips in the moderator flux become very evident. Intermediate cases
can be approximated[55] by assuming

ϕm = (1 − β) C(E) + βϕf , (273)

where β is a heterogeneity parameter given by

Vf σe
β= . (274)
Vm σm
Note that β → 0 gives the isolated rod limit and β → 1 gives the close-packed
lattice limit. This substitution reduces the calculation of the fuel flux to

(σf + σe ) ϕf = (1 − β) C(E) σe + Sβ , (275)

where Sβ is the source term corresponding to a homogeneous mixture of the

fuel isotopes with the isotopes from the moderator region changed by the factor
βσe /σm . If the fuel and moderator each consisted of a single isotope and for
isotropic scattering in the center-of-mass system, the integral equation would
become

[σ0 + σt (E)] ϕf (E) = (1 − β) C(E) σ0

Z E/αm
βσ0
+ ϕ (E ′ ) dE ′
′ f
E (1 − α m )E
σsf (E ′ )
Z E/αf
+ ϕ (E ′ ) dE ′ ,
′ f
(276)
E (1 − α f )E

where σ0 is σe divided by the fuel density (units are barns/atom), αm and αf are
the maximum fractional energy change in scattering for the two isotopes, and
σsf (E ′ ) is the fuel scattering cross section.
This result has a form parallel to that of Eq. 272, but the solution depends
on the two parameters β and σ0 . For any given data set, β must be chosen in
advance. This might not be difficult if the data are to be used for one particular
system, such as pressurized water reactors. The routine also has the capability
to include one more moderator integral with a different α value and a constant
cross section. The full equation is

NJOY2016 199
LA-UR-17-20093 8 GROUPR

[σ0 + σt (E)] ϕf (E) = (1 − β) C(E) σ0

Z E/α3
β(1 − γ)(σ0 − σam
+ ′
ϕf (E ′ ) dE ′
E (1 − α 3 )E
Z E/α2
σam + βγ(σ0 − σam
+ ′
ϕf (E ′ ) dE ′
E (1 − α 2 )E
σsf (E ′ )
Z E/αf
+ ϕ (E ′ ) dE ′ ,
′ f
(277)
E (1 − α f )E

where σam is the cross section of the admixed moderator (with energy loss α2 ),
and γ is the fraction of the admixed moderator that is mixed with the external
moderator (which has energy loss α3 ). This allows calculations with H2 O as
the moderator and an oxide as the fuel. The flux calculator can thus obtain
quite realistic flux shapes for a variety of fuel, admixed moderator and external
moderator combinations. An example comparing the Bondarenko flux with a
more realistic computed flux is given in Fig. 25.

Figure 25: A comparison of the Bondarenko flux model (dashed) with a realistic computed
flux (solid) for a 238 U oxide pin in water in the region of the 6.7 eV resonance.

In practice, a fuel rod rarely contains only one resonance isotope. As an

example, consider a mixture of a few percent of 239 Pu with 238 U as the major
component. There will be a strong dip in the flux associated with the 6.7 eV 238 U
resonance that will affect the flux in the region of the 7.8 eV 239 Pu resonance
(the interference effect), and there will also be a dip in the flux corresponding
to the 7.8 eV resonance (the self-shielding effect). This additional complication

200 NJOY2016
8 GROUPR LA-UR-17-20093

in the flux shape would be expected to change the group constants for 239 Pu

since both features lie in the same group for typical group structures. However,
the effect of the 239 Pu on the 238 U group constants should be minimal. This
argument suggests that the full flux calculation be used for 238 U as a single
resonance material. The resulting flux would then be used to estimate the flux
to be used in averaging the 239 Pu cross sections as follows:

ϕ238 (E, 50)

ϕ239 (E, σ0 ) = , (278)
σ239 (E)
σ0 +
σ238 (E)
1+
50

where the 238 U flux is characteristic of a background of 50 barns/atom, which is

representative of many thermal reactor systems. This formula assumes that the
effect of 239 Pu on the scattering source for the mixture is small, but it retains the
absorption effects. The self-shielding of 239 Pu is treated in the narrow resonance
approximation only. The GROUPR flux calculator includes an option to write
out a file containing the calculated flux and cross section needed for this formula
(e.g., for 238 U) and another option to skip the flux calculation and use the formula
above to obtain the weighting flux (e.g., for 239 Pu).

The slowing-down integral equation of Eq. 272 or 276 is solved point by point
(see subroutine genflx) using the total and elastic cross sections on the PENDF
tape produced by RECONR. In order to keep this task within bounds, the flux
is computed from the lower limit of the first group to a specified energy fehi or
until nfmax values have been computed. The flux at higher energies is continued
using the Bondarenko model described above.

8.6 Fission Source

The fission source was included in the transfer cross section σX in the develop-
ment above. It is usually convenient to separate fission from scattering. Assum-
ing isotropy, the source of fission neutrons into group g is given by

X
Sg = σf g′ →g ϕ0g′ , (279)
g′

where the group-to-group matrix for fission is defined as in Eq. 251, but with
ℓ equal to zero. Most existing transport codes do not use this matrix form
directly because the upscatter is expensive to handle and a reasonably accurate
alternative exists. Except for relatively high neutron energies, the spectrum of

NJOY2016 201
LA-UR-17-20093 8 GROUPR

fission neutrons is only weakly dependent on initial energy. Therefore, the fission
source can be written

X
Sg = χg ν g′ σf g′ ϕ0g′ , (280)
g′

where ν g is the fission neutron yield, σf g is the fission cross section, and χg is
the average fission spectrum (the familiar “chi” vector), which can be defined by
X
σf g′ →g ϕ0g′
g′
χg = X X . (281)
σf g′ →g ϕ0g′
g g′

The fission neutron yield is given by

X
νg = σf g→g′ / σf g . (282)
g′

Clearly, χg as given by Eq. 281 depends on the flux in the system of interest.
The dependence is weak except for high incident energies, and a rough guess for
ϕ0g usually gives an accurate spectrum. When this is not the case, the problem
can be iterated, or the full matrix representation can be used.
It is possible to take advantage of the weak energy dependence of the shape of
the fission spectrum at low energies to reduce the time required to process fission
data, and to reduce the size of the fission data on the output file. GROUPR
determines a break energy from File 5 such that the fission spectrum is constant
below this energy. It only has to make a single calculation of this spectrum χLE
g .

P g for groups
Then it computes a fission neutron production cross section σfHE
below the break energy using the normal cross section processing methods. A
full matrix σfHE
g ′ →g is computed for groups above the break point. As an example,
consider using the GROUPR 187-group structure and finding that there are
130 constant groups. The 187×187 matrix is reduced to a 57×187 matrix, a
187-element spectrum vector, and a 130-group production vector, for a total
reduction in size of 68%. The effective fission matrix is given by

σf g′ →g = σfHE LE LE
g ′ →g + χg σf P g ′ . (283)

The fission matrix computed by GROUPR represents the prompt part of fis-
sion only. The delayed component of fission is represented by a delayed-neutron

202 NJOY2016
8 GROUPR LA-UR-17-20093

yield ν D
g , decay constants for six (or sometimes 8) time groups, λi , and emission
D

probability spectra for six (or 8) time groups, χD

ig . Steady-state fission can be
obtained using

ν SS D
g = νg + νg , (284)

and
X X
σf g′ →g ϕ0g′ + χD
g νD
g ′ σf g ′ ϕ0g ′
g′ g′
χSS
g = XX X , (285)
σf g′ →g ϕ0g′ + νD
g ′ σf g ′ ϕ0g ′
g g′ g′

where

X
χD
g = χD
ig . (286)
i

Note that χD
g sums to unity, but the χ for each time group sums to the fraction
of the delayed neutron yield that appears in that time group.
In ENDF-format files, the total fission reaction is represented by MT=18.
Important isotopes also give the partial fission reactions (n,f), (n,n′ f), (n,2nf),
and sometimes (n,3nf) using MT=19, 20, 21, and 38 respectively. The MT=18
representation is adequate for most fission reactor applications, but the partial
reactions should be processed for applications with significant flux above 6 MeV.
Caution: although the cross section for MT=18 equals the sum of its parts,
the group-to-group fission matrix σf g→g′ computed from MT=18 will not, in
general, equal the sum of the partial matrices for MT=19, 20, 21, and 38 above
the 6-MeV threshold for second-chance fission. The breakup into partial fission
matrices has not been used in recent ENDF/B-VII evaluations. The delayed
neutron data are given in MT=455. Sample input instructions for processing
the various combinations of fission reactions used in ENDF/B will be found in
Section 8.14. GROUPR outputs all the components of fission separately in order
to give succeeding modules or codes complete flexibility.

8.7 Diffusion Cross Sections

The diffusion equation is often used in reactor physics calculations. Starting
with Eq. 246, use the Legendre expansion for ϕg in the derivative term, and
make use of the recursion relation for µPℓ (µ) and the orthogonality relation for
the Legendre polynomials to obtain the transport equation in spherical-harmonic

NJOY2016 203
LA-UR-17-20093 8 GROUPR

form

n + 1 ∂ϕn+1,g n ∂ϕn−1,g
+ σtng ϕng
2n + 1 ∂x 2n + 1 ∂x
X
= σsng′ →g ϕng′ + Sf g + Qng , (287)
g′

where the transfer term has been separated into a scattering term with cross
section σs , and a fission source term Sf . When this set of equations is truncated
at n = N, the results are usually called the “PN equations”. For now, all terms
with n > 1 are dropped, and Q is assumed to be isotropic. Thus,

∂ϕ1g X
+ σt0g ϕ0g = σs0g′ →g ϕ0g′ + Sf g + Q0g , (288)
∂x ′ g

and

1 ∂ϕ0g X
+ σt1g ϕ1g = σs1g′ →g ϕ1g′ . (289)
3 ∂x ′ g

The second equation can be written in the form of Fick’s Law as follows:

∂ϕ0g
ϕ1g = −Dg , (290)
∂x

1 1
Dg = X , (291)
3 σt1g − σs1g′ →g ϕ1g′ / ϕ1g
g′

where Dg is the diffusion constant. The term in the denominator of the second
factor is the transport cross section for diffusion, σtrD . Unfortunately, it depends
on a fairly complete knowledge of the neutron current in the system, perhaps
from a previous calculation. However, for many problems, σtrD can be simplified
by assuming that

X X
σs1g′ →g ϕ1g′ ≈ σs1g→g′ ϕ1g , (292)
g′ g′

with the result that

X
σtrD,g = σt1g − σs1g→g′ , (293)
g′

204 NJOY2016
8 GROUPR LA-UR-17-20093

or

σtrD,g = σt1g − µg σs0g , (294)

where µg is the average scattering cosine for neutrons in group g. These forms
depend only on the shape of the weighting flux within the group, as usual.
Substituting for ϕ1g in Eq. 288 gives

∂ ∂ϕ0g
X
−Dg + σ0tg ϕ0g = σs0g′ →g ϕ0g′ + Sf g + Q0g , (295)
∂x ∂x ′ g

which is the standard diffusion equation in slab geometry. Neither the diffusion
coefficient nor the transport cross section for diffusion is produced directly by
GROUPR. However, the components such a σtℓ and σs0g′ →g are made available
to subsequent modules.

8.8 Cross Sections for Transport Theory

The SN (discrete ordinates) transport codes solve the equation

∂
µ ϕg (µ, x) + σgSN ϕg (µ, x)
∂x
N
X 2ℓ + 1 X
SN
= Pℓ (µ) σsℓg ′ →g (x) ϕℓg ′
2 ′
ℓ=0 g
+ Sf g + Qg (µ, x) , (296)

where once again one-dimensional slab geometry has been used for simplicity.7
By comparing Eq. 296 with Eq. 246, it is seen that the SN equations require the
following cross sections:

SN
σsℓg ′ →g = σsℓg ′ →g , g ′ ̸= g , (297)

and

SN
σsℓg→g = σsℓg→g − σtℓg + σgSN , (298)

where σgSN is not determined and can be chosen to improve the convergence
of the SN calculation. A particular choice of σgSN gives rise to a “transport
approximation”, and various recipes are in use, such as:
7
The following development is based on the work of Bell, Hansen, and Sandmeier[56].

NJOY2016 205
LA-UR-17-20093 8 GROUPR

Consistent-P approximation:

σgSN = σt0g . (299)

Inconsistent-P approximation:

σgSN = σt,N +1,g . (300)

Diagonal transport approximation:

σgSN = σt,N +1,g − σs,N +1,g→g . (301)

Bell-Hansen-Sandmeier (BHS) or extended transport approximation:

X
σgSN = σt,N +1,g − σs,N +1,g→g′ . (302)
g′

Inflow transport approximation:

X
σs,N +1,g′ →g ϕN +1,g′
g′
σgSN = σt,N +1,g − . (303)
ϕN +1,g
The first two approximations are most appropriate when the scattering orders
above N are small. The inconsistent option removes most of the delta function
of forward scattering introduced by the correction for the anisotropy of the total
scattering rate, and should normally be more convergent than the consistent
option. The diagonal and BHS recipes make an attempt to correct for anisotropy
in the scattering matrix and are especially effective for forward-peaked scattering.
The BHS form is more often used, but the diagonal option can be substituted
when BHS produces negative values. The inflow recipe makes the N +1 term
of the PN expansion vanish, but it requires a good knowledge of the N +1 flux
moment from some previous calculation. Inflow reduces to BHS for systems
in equilibrium by detail balance (i.e., the thermal region). In the absence of
self-shielding (that is, σ0 →∞), the distinction between σt1 and σt0 disappears,
and so does the distinction between the inconsistent-P and consistent-P options.
Also note that the inflow and BHS versions of σg are equivalent to the definitions
of σtrD,g given in Eqs. 291 and 293, respectively, when N =0.
The transport-corrected cross sections are not computed directly by GROUPR,
but the components needed are written to the GROUPR output file for use by
subsequent modules.

206 NJOY2016
8 GROUPR LA-UR-17-20093

8.9 Photon Production and Coupled Sets

Photon transport is treated with an equation similar to Eq. 246, except that
the flux, cross sections, and groups are all referred to photon interactions and
photon energies instead of to the corresponding neutron quantities. Methods
of producing the photon interaction cross sections are described in GAMINR.
The external photon source Qg depends on the neutron flux and the photon
production cross sections through

∞
X 2ℓ + 1 X
Qg (x, µ) = Pℓ (µ) σγℓg′ →g (x) ϕℓg′ (x) , (304)
2
ℓ=0 g′

where σγℓg′ →g is defined by Eq. 251 with X replaced by γ. The ENDF files
define σγ using a combination of photon production cross sections (MF=13),
photon yields (MF=12) with respect to neutron cross sections (MF=3), discrete
lines (MF=12 and 13), and continuous γ distributions (MF=15). Methods for
working with these representations will be discussed in more detail below.
The low-energy groups for fission and capture normally have photon emission
spectra whose shapes do not change with energy. The same method used for
reducing the size of the fission matrix (see Section 8.6) can be used for these
photon production matrices. In mathematical form,

HE
σγg′ →g = σγg LE LE
′ →g + sγ gσγP g ′ , (305)

where sLE
γ g is the normalized emission spectrum, and σγP g is the associated
LE

production cross section.

For many practical problems, it is convenient to combine the neutron and
photon transport calculations into a single application of Eq. 246 where the pho-
tons are treated as additional groups of low-energy neutrons. Since (γ,n) events
are not usually very important, the downscattering structure (see Section 8.2) of
the transport calculation is preserved for both n→γ and n→n events. Cross sec-
tions for this application are called “coupled sets”. Coupled sets are not produced
directly by GROUPR, but the n→n and n→γ components are made available
to the other modules where they can be combined with γ→γ cross sections
from GAMINR. As an example, the MATXSR module can format data for the
TRANSX code, which can, in turn, prepare coupled sets for use by standard
transport codes such as ONEDANT[14] or PARTISN[17]. The normal arrange-

NJOY2016 207
LA-UR-17-20093 8 GROUPR

ment of data in a coupled set is shown schematically in Fig. 26.

8.10 Thermal Data

GROUPR uses the thermal data written onto the PENDF tape by the THERMR
module. It does not process ENDF File-7 data directly. Three different repre-
sentations of thermal scattering are used in ENDF.
For crystalline materials like graphite, coherent elastic (with zero energy
change) scattering can take place. The cross section for this process shows well-
defined Bragg edges at energies that correspond to the various lattice-plane spac-
ings in the crystalline powder. As shown in THERMR, the angular dependence
of the coherent elastic cross section can be written as

edits

σa,νσf,σt

γ→ n 0

n→n, up γ→ γ, up

typical γ→ γ, down
truncated
upscatter n→ n, down

n→ γ

n γ

Figure 26: Arrangement of neutron (n) and photon (γ) cross sections in a coupled trans-
port table. The group index increases from left to right, and the position index
increases from top to bottom. The γ→n and γ→γ upscatter blocks are normally
empty.

208 NJOY2016
8 GROUPR LA-UR-17-20093

X bi
σ coh (E, µ) = δ(µ − µi ) , (306)
E
i

where

Ei
µi = 1 − 2 , (307)
E
and where the Ei are the energies of the Bragg edges. THERMR integrates
Eq. 306 over all angles, and writes the result to the PENDF tape. Clearly, the
bi can be recovered from

X′
E σ coh (E) = bi , (308)
i

where the primed sum is over all i such that Ei < E. It is only necessary to
locate the steps in E σ coh (E). The size of the step gives bi , and the E for the
step gives Ei . The Legendre cross sections become

X′ bi
σℓcoh (E) = Pℓ (µi ) , (309)
E
i

where any terms with µi < − 1 are omitted from the primed sum. An example
of a pointwise cross section for coherent elastic scattering is given in Fig. 27.

Figure 27: The coherent elastic scattering cross section for BeO showing the Bragg edges.
The shape of σ(E) between edges is 1/E. Therefore, the function Eσ(E) is a
stair-step function, where the height of each step depends on the structure factor
for scattering from that set of lattice planes (see Eq. 308).

NJOY2016 209
LA-UR-17-20093 8 GROUPR

For hydrogenous solids like polyethylene and zirconium hydride, the process
of incoherent elastic scattering is important. Here the angular cross section is
given by
 
σb 2EW
iel
σ (E, µ) = exp − (1 − µ) . (310)
2 A

THERMR converts this into an integrated cross section, σ iel (E), and a set of N
equally probable emission cosines, µi . These angles are present in File 6 on the
PENDF tape. GROUPR can easily determine the Legendre components of the
scattering cross section using

N
1 X
σℓiel (E) = σ iel (E) Pℓ (µi ) . (311)
N
i=1

The third thermal process is incoherent inelastic scattering. Here the neutron
energy can either increase or decrease. The data from THERMR are given as
a cross section in File 3 and an energy-angle distribution using a special form
of File 6. The distribution is represented by sets of secondary-energy values
E ′ for particular incident energies E. For each E→E ′ , a scattering probability
f inc (E→E ′ ) and a set of equally probable cosines µi (E→E ′ ) are given. The
scattering probabilities for each value of E integrate to unity. Although the
thermal scattering cross section is a smooth function of incident neutron energy,
this is not true for the scattering from E to one particular final energy group
g ′ , since the differential cross section tends to peak along the line E ′ =E and at
energy-transfer values corresponding to well-defined excitations in the molecule
or lattice. If interpolation between adjacent values of E were to be performed
along lines of constant E ′ , the excitation peaks and the E=E ′ feature would
produce double features in the intermediate spectrum, as shown in Fig. 28. To
avoid this problem, while still using a relatively sparse incident energy grid,
GROUPR interpolates between E and E ′ along lines of constant energy transfer.
Of course, this breaks down at low values of E ′ , because one of the spectra will go
to zero before the other one does. In this range, GROUPR transforms the low-
energy parts of the two spectra onto a “unit base,” combines them in fractions
that depend on E, and scales the result back out to the interpolated value of E ′
corresponding to E.

210 NJOY2016
8 GROUPR LA-UR-17-20093

Ehi

Emid

Elo

E'

Figure 28: Illustration of thermal interpolation showing the double-humped curve resulting
from simple Cartesian interpolation for a discrete excitation (solid) and the more
realistic curve obtained by interpolating along lines of constant energy transfer
(dashed).

8.11 Generalized Group Integrals

In order to unify many formerly different processing tasks, GROUPR uses the
concept of a generalized group integral
Z
F(E) σ(E) ϕ(E) dE
g
σg = Z , (312)
ϕ(E) dE
g

where the integrals are over all initial neutron energies in group g, σ(E) is a cross
section at E, and ϕ(E) is an estimate of the flux at E. The function F(E) is
called the “feed function”. It alone changes for different data types. To average
a neutron cross section, F is set to 1. To average a ratio quantity like µ with
respect to elastic scattering, F is set to µ. For photon production, F is the
photon yield. For matrices, F is the ℓ-th Legendre component of the normalized
probability of scattering into secondary energy group g ′ from initial energy E.
This definition is clearly independent of whether the secondary particle is a
neutron or a photon.
The question of integration grid or quadrature scheme is important for the

NJOY2016 211
LA-UR-17-20093 8 GROUPR

evaluation of Eq. 312. Each factor in the integrands has its own characteristic
features, and it is important to account for them all. First, a grid must be
established for each factor. As an example, the grid of σ(E) is generated in
RECONR such that sigma can be obtained to within a given tolerance by linear
interpolation. GROUPR contains a subroutine getsig which carries out this
interpolation at E and also returns the next grid energy in enext. Subroutines
getflx and getff perform similar functions for the flux and feed function. It
is now easy to generate a union grid for the three-factor integrand using the
following Fortran:

...
call getsig(e,enext,...)
call getflx(e,en,...)
if (en.lt.enext) enext=en
call getff(e,en,...)
if (en.lt.enext) enext=en
...

On occasion, there will be a discontinuity at enext for one of the factors. In

order to flag such a case, each “get” routine also sets a discontinuity flag idisc.
The grid logic actually used throughout NJOY is as follows:

...
call getsig(e,enext,idisc,...)
call getflx(e,en,idis,...)
if (en.eq.enext.and.idis.gt.idisc) idisc=idis
if (en.lt.enext) idisc=idis
if (en.lt.enext) enext=en
call getff(e,en,idis,...)
if (en.eq.enext.and.idis.gt.idisc) idisc=idis
if (en.lt.enext) idisc=idis
if (en.lt.enext) enext=en
...

This union grid for the integrand in the numerator is used to subdivide the gen-
eralized group integral of Eq. 312 into “panels”. The main program of GROUPR
carries out the integrals with the following logic:

...
elo=egn(ig)
ehi=egn(ig+1)
enext=ehi

212 NJOY2016
8 GROUPR LA-UR-17-20093

230 call panel(elo,enext,...)

if (enext.eq.ehi) go to 240
elo=enext
enext=ehi
go to 230
240 continue
...

A panel is first defined by the energy bounds of group g. Subroutine panel is

called to sum in the contributions from this panel. However, panel discovers
that the integrand has a grid point at enext less than ehi. It adds in the
contributions for the smaller panel elo to enext and returns. GROUPR now
sees that enext is less than ehi, so it tries a new panel from the top of the last
panel (elo=enext) to ehi. This loop continues until a panel with ehi as its
upper bound is summed in. The integral for this group is then complete.
In this simple way, the algorithm accounts for the user’s group structure and
for all structure in the integrand. The method used for establishing the RECONR
grid makes this integration algorithm equivalent to adaptive integration as used
in MINX[21]. It has the great advantage that no “stack” of intermediate results
is carried along. This single-pass feature of the quadrature scheme allows many
different integrals to be accumulated simultaneously within reasonable storage
limits. In this way, GROUPR accumulates cross sections for all values of σ0
simultaneously. Similarly, group-to-group cross sections are computed for all
secondary energy groups and all Legendre orders simultaneously.
Any degree of complexity can be used for the integral over each subpanel. Be-
cause σ(E) has been linearized, panel is based on trapezoidal integration. Both
ϕ(E) and R(E) = σ(E)×ϕ(E) are assumed to vary linearly across each panel.
In some cases, the feed function is oscillatory over a certain energy range (see
Two-Body Scattering, Section 8.12). It is then convenient to integrate inside the
panel using Lobatto quadrature[57] (note the variable nq in panel). As discussed
in more detail later, this method can obtain accurate results for an oscillatory
function whose integral is small with respect to its magnitude. This behavior is
characteristic of the higher Legendre components of two-body scattering cross
sections.
The generalized integration scheme discussed here is also used by the GAM-
INR and ERRORR modules.

NJOY2016 213
LA-UR-17-20093 8 GROUPR

8.12 Two-Body Scattering

Elastic scattering (ENDF/B MT=2) and discrete inelastic neutron scattering
(with MT=51 – 90) are both examples of two-body kinematics and are treated
together by GROUPR. Some other cases that occur for charged particles or in
File 6 will be discussed later. The feed function required for the group-to-group
matrix calculation may be written
Z Z +1
′
Fℓg′ (E) = dE dω f (E→E ′ , ω) Pℓ (µ[ω]) , (313)
g′ −1

where f (E→E ′ , ω) is the probability of scattering from E to E ′ through a center-

of-mass cosine ω and Pℓ is a Legendre polynomial for laboratory cosine µ. The
laboratory cosine corresponding to ω is given by

1 + Rω
µ= √ , (314)
1 + R2 + 2Rω
and the cosine ω is related to secondary energy E ′ by

E ′ (1 + A)2 /A′ − E(1 + R2 )

ω= , (315)
2RE
where A′ is the ratio of the emitted particle mass to the incident particle mass
(A′ =1 for neutron scattering). In Eqs. 314 and 315, R is the effective mass ratio
r r
A(A + 1 − A′ ) (A + 1)(−Q)
R= 1− , (316)
A′ AE
where A is the ratio of target mass to neutron mass, and −Q is the energy level
of the excited nucleus (Q=0 for elastic scattering). Integrating Eq. 313 over
secondary energy gives
Z ω2
Fℓg′ (E) = f (E, ω) Pℓ (µ[ω]) dω , (317)
ω1

where ω1 and ω2 are evaluated using Eq. 315 for E ′ equal to the upper and lower
bounds of g ′ , respectively. The scattering probability is given by

L
X
f (E, ω) = fℓ (E) Pℓ (ω) , (318)
ℓ=0

where the Legendre coefficients are either retrieved directly from the ENDF File 4
or computed from File 4 tabulated angular distributions (see subroutines getfle
and getco).

214 NJOY2016
8 GROUPR LA-UR-17-20093

The integration in Eq. 317 is performed (see subroutine getdis) simultane-

ously for all Legendre components using Gaussian quadrature[57]. The quadra-
ture order is selected based on the estimated polynomial order of the integrand.
A reasonable estimate is given by

ND + NL + log(300/A) , (319)

where ND is the number of Legendre components desired for the feed function,
and NL is the number of components required to represent f (E, ω). The log
term approximates the number of additional components required to represent
the center-of-mass to lab transformation.
The two-body feed function for higher Legendre orders is a strongly oscillatory
function in some energy ranges. An example is shown in Fig. 29. Furthermore,
the integral of the oscillatory part is often small with respect to the magnitude
of the function. Such functions are very difficult to integrate with adaptive tech-
niques, which converge to some fraction of the integral of the absolute value.
This is the reason that MINX[21] gave poor answers for small Legendre com-
ponents of the scattering matrix. Gaussian methods, on the other hand, are
capable of integrating such oscillatory functions exactly if they are polynomials.
Since a polynomial representation of the feed function is fairly accurate, a Gaus-
sian quadrature scheme was chosen. The scheme used is also well adapted to
performing many integrals in parallel. In GROUPR, all Legendre components
and all final groups are accumulated simultaneously (see panel).
The boundaries of the various regions of the feed function are called “critical
points.” Between critical points, the feed function is a smooth analytic function
of approximately known polynomial order. It is only necessary to add these
critical points to the incident energy grid of the feed function (the enext variable)
and to tell panel what quadrature order (nq) to use. The critical points are
determined in getff by solving Eq. 315 for the values of E for which ω = +1
and ω = −1 when E ′ is equal to the various group boundaries. This can be done
by writing

Eg 1
(320)
 2 
= 2
R ± 2R + 1 ,
E (1 + A)

substituting for E using Eq. 316, and then solving for R. The result is

NJOY2016 215
LA-UR-17-20093 8 GROUPR

Figure 29: A typical feed function for two-body scattering showing the oscillations that must
be treated correctly by the integration over incident energy.

(A + 1)(−Q)
Ecrit = A , (321)
A′ F 2
1−
A(A + 1 − A′ )

where √
1± D
F = , (322)
EF
1+
(−Q)

A(A + 1 − A′ )
   
EF EF
D= ′
1+ −1 , (323)
A (−Q) (−Q)

and
1+A
EF = Eg . (324)
A + 1 − A′
File 4 can also contain angular distributions for charged-particle emission
through discrete levels (for ENDF/B-VI and later see MT=600 – 648 for protons,
MT=650 – 698 for deuterons, and so on; the elastic case, MT=2, is discussed
in the next section). Moreover, File 6 can contain angular distributions for
discrete two-body scattering (see Law 3). It can also declare that a particular
particle is the recoil particle from a two-body reaction (Law 4), in which case
the appropriate angular distribution is obtained from the corresponding Law

216 NJOY2016
8 GROUPR LA-UR-17-20093

3 subsection by complementing the angle. The representation of the angular

distribution for these laws is almost identical to that in File 4, and the calculation
is done in getdis using much of the File 4 coding. The effects on the kinematics
of the difference in the mass of the emitted and incident particles are handled
by the variable A′ in the above equations.

8.13 Charged-Particle Elastic Scattering

Coulomb scattering only occurs in the elastic channel, and this calculation also
uses the getdis subroutine. The problem with Coulomb scattering is that the
basic Rutherford formula becomes singular at small angles. In practice, this
singularity is removed by the screening effects of the atomic electrons. The for-
ward transport of charged particles in this screening regime is usually handled by
continuous-slowing-down theory by using a “stopping power”. The new ENDF-6
format allows for three different ways to handle the large-angle scattering regime.
First and most general is the nuclear amplitude expansion:

σ = |nucl + coul|2
= σnucl + σcoul + interference (325)

The Coulomb term is given by the Rutherford formula, a Legendre expansion

is defined for the nuclear term, and a complex Legendre expansion is defined
for the interference term. This representation cannot be generated directly from
experimental data; an R-matrix or phase-shift analysis is necessary.
A method very closely related to experiment (σexp ) is the nuclear plus inter-
ference (NI) formula:

σNI (µ, E) = σexp (µ, E) − σcoul (µ, E), (326)

where the function is only defined for angles with cosines µ<µmax . The mini-
mum angle is usually taken to be somewhere around 20 degrees (GROUPR uses
µmax =0.96). This function is still ill-behaved near the cutoff, and it must be
tabulated. The third option is the residual cross section expansion:

(327)
 
σR (µ, E) = (1 − µ) σexp (µ, E) − σcoul (µ, E) .

The (1 − µ) term removes the pole at the origin. The residual is uncertain, but
it is usually small enough that the entire curve can be fitted with Legendre poly-

NJOY2016 217
LA-UR-17-20093 8 GROUPR

nomials without worrying about what happens at small angles. In practice, both
the nuclear amplitude expansion and the nuclear plus interference representation
are converted to the residual cross section representation in subroutine conver.
As a result, getdis only has to cope with the one representation.

8.14 Continuum Scattering and Fission

In ENDF evaluations, scattering from many closely-spaced levels or multibody
scattering such as (n,2n), (n,n′ α) or fission is often represented using a separable
function of scattering cosine and secondary neutron energy

f (E→E ′ , µ) = F (E, µ) g(E→E ′ ) , (328)

where F is the probability that a neutron will scatter through a laboratory angle
with cosine µ irrespective of final energy E ′ . It is obtained from MF=4. Similarly,
g is the probability that a neutron’s energy will change from E to E ′ irrespective
of the scattering angle, and it is given in MF=5. Continuum reactions are mostly
identified by MT numbers of 6 – 49 and 91. Recently, previously unused MT
numbers, 152 through 200, were assigned to additional continuum reactions that
are beginning to appear in specialized evaluated files that extend beyond 20 MeV
(e.g., the International Reactor Dosimetry and Fusion File). Secondary-energy
distributions, whether found in MF=5 or MF=6 are represented by “Laws" as
follows:
Law Description
1 Arbitrary tabulated function
5 General evaporation spectrum
(Used for delayed neutrons only.)
7 Simple Maxwellian fission spectrum
9 Evaporation spectrum
11 Energy-dependent Watt fission spectrum
12 Energy-dependent spectrum of Madland and Nix

The feed functions for continuum scattering are simply

Z +1 Z
′
Fℓg′ (E ) = Pℓ (µ) f (E, µ) dµ g(E, E ′ ) dE ′ . (329)
−1 g′

The first integral is returned by getfle [“fle” for fℓ (E)] as described above, and

218 NJOY2016
8 GROUPR LA-UR-17-20093

the second integral is returned by getsed (“sed” for secondary-energy distribu-

tion).
For Law 1, g is given as a tabulated function of E ′ for various values of E.
When E1 ≤ E < E2 , the term g′ g(E, E ′ ) dE ′ is obtained by interpolating be-
R

tween precomputed values of g′ g(E1 , E ′ ) dE ′ and g′ g(E2 , E ′ ) dE ′ in subroutine

R R

getsed. Except in the case of fission, any apparent upscatter produced by the
“stairstep” treatment near E=E ′ is added to the in-group scattering term (g ′ =g).
For Law 5, g(x) is given versus x = E ′ /θ(E) and θ(E) is given vs. E in File
5. This secondary neutron distribution leads to the following group integral:
Z Z E2 /θ
g(E, E ′ ) dE ′ = θ(E) g(x) dx , (330)
g′ E1 /θ

with E1 and E2 being the lower and upper boundary energies for group g ′ .
For Law 7, the secondary-energy distribution is given by
√
E′ E′
 
′
g(E, E ) = exp − , (331)
I θ(E)

where the effective temperature θ(E) is tabulated in File 5 and the normalization
factor is given by
r
π
3/2
erf(x) − xe−x , (332)
 
I=θ
4

where

E−U
x= . (333)
θ

Here U is a constant used to define the upper limit of secondary neutron energy
θ ≤ E ′ ≤ E−U . The desired group integral is given by

Z Z E2
g(E, E ′ ) dE ′ = g(E, E ′ ) dE ′
g′ E1
X1 − X2 − Y1 + Y2
= p , (334)
π/4 − Y − X

where

NJOY2016 219
LA-UR-17-20093 8 GROUPR

√
X= x e−x , (335)

√
r
π
Y = rerfc( x) , (336)
4
and where X1 , Y1 , X2 , and Y2 refer to X and Y evaluated at E1 /θ and E2 /θ.
The integral of Eq. 334 is computed in anased (“ana” for analytic). The function
“rerfc” is the reduced complementary error function[57].
For Law 9, the secondary-energy distribution is given by

E′ E′
 
′
g(E, E ) = exp − , (337)
I θ(E)
where

I = θ2 1 − e−x (1 + x) . (338)
 

Here x has the same meaning as above – see Eq. 333. The group integral is

(1 + x1 )e−x1 − (1 + x2 )e−x2
Z
g(E, E ′ ) dE ′ = , (339)
g′ 1 − (1 + x)e−x
where x1 and x2 refer to E1 /θ and E2 /θ, respectively. This result is also com-
puted in anased.
For Law 11, the secondary-energy distribution is given by
′
′ e−E /a √

g(E, E ) = sinh bE ′ , (340)
I
where

r
1 πa3 b x0  √ √ √ √ 
I= e erf( x − x0 ) + erf( x + x0 ) − ae−x sinh(abx) . (341)
2 4

In this case, a(E) and b(E) are given in File 5,

E−U
x= , (342)
a

and

ab
x0 = . (343)
4

220 NJOY2016
8 GROUPR LA-UR-17-20093

The group integrals are given by

Z
H(x1 , x2 , x0 )
g(E, E ′ ) = , (344)
g′ H(0, x, x0 )

where

√ √ √ √
H(x1 , x2 , x0 ) = H2 ( x1 − x0 , x2 − x0 )
√ √ √ √ √
+ x0 H1 ( x1 − x0 , x2 − x0 )
√ √ √ √
− H2 ( x1 + x0 , x2 + x0 )
√ √ √ √ √
+ x0 H1 ( x1 + x0 , x2 + x0 ) , (345)

and where
Z b
1 2
Hn (a, b) = z n e−z dz . (346)
π a

√
The methods for computing Hn are described in BROADR. When x2 < .01, a
short-cut calculation can be used for the numerator of Eq. 344
√
4 x0 e−x0  3/2 3/2 
p x2 − x1 . (347)
3 π/4

For Law 12, the Madland-Nix option, the secondary-energy distribution is

given by

1
g(E, E ′ ) = G(E ′ , Ef l ) + G(E ′ , Ef h ) , (348)

2
where

G(E ′ , Ef )
 
1 3/2 3/2
= p u E1 (u2 ) − u1 E1 (u1 ) + γ(3/2, u2 ) − γ(3/2, u1 ) ,
3 Ef Tm 2
(349)

NJOY2016 221
LA-UR-17-20093 8 GROUPR

√
u1 = ( E ′ − Ef )2 /Tm , and (350)
p

√
u2 = ( E ′ + Ef )2 /Tm , (351)
p

and where Ef l , Ef h , and Tm (E) are given in File 5. The special functions
used are the first-order exponential integral, E1 (x), and the incomplete gamma
function, γ(n, x). The group integrals of this function are very complex[58]. Let

p
α = Tm , (352)
(353)
p
β = Ef ,
√
( E1 + β)2
A = 2
, (354)
√ α
( E2 + β)2
B = 2
, (355)
√ α
( E1 − β)2
A′ = 2
, and (356)
√ α
( E2 − β)2
B′ = . (357)
α2

Then the integral over the range (E1 , E2 ) of G is given by one of the following
three expression, depending on the region of integration in which E1 and E2 lie.
Region I (E1 ≥ Ef , E2 > Ef )
Z E2
G(E ′ , Ef ) dE ′
p
3 Ef Tm
E1
 2 1 2 1
= ( α2 B 5/2 − αβB 2 )E1 (B) − ( α2 A5/2 − αβA2 )E1 (A)

5 2 5 2
 2 5/2 1 2 2 5/2 1 2
− ( α2 B ′ + αβB ′ )E1 (B ′ ) − ( α2 A′ + αβA′ )E1 (A′ )

5 2 5 2
+ (α2 B − 2αβB 1/2 ) γ(3/2, B) − (α2 A − 2αβA1/2 ) γ(3/2, A)
 

1/2 1/2
− (α2 B ′ + 2αβB ′ γ(3/2, B ′ ) − (α2 A′ + 2αβA′ ) γ(3/2, A′ )
 

3 
− α2 γ(5/2, B) − γ(5/2, A) − γ(5/2, B ′ ) + γ(5/2, A′ )

5
3 ′ ′
− αβ e−B (1 + B) − e−A (1 + A) + e−B (1 + B ′ ) − e−A (1 + A′ ) .
 
2
(358)

222 NJOY2016
8 GROUPR LA-UR-17-20093

Region II (E1 < Ef , E2 ≤ Ef )

Z E2
G(E ′ , Ef ) dE ′
p
3 (Ef Tm )
E1
 2 1 2 1
= ( α2 B 5/2 − αβB 2 )E1 (B) − ( α2 A5/2 − αβA2 )E1 (A)

5 2 5 2
 2 2 2 5/2 1 2
− ( α2 B ′ )E1 (B ′ ) − ( α2 A′ − αβA′ )E1 (A′ )

5 5 2
+ (α2 B − 2αβB 1/2 ) γ(3/2, B) − (α2 A − 2αβA1/2 ) γ(3/2, A)
 

1/2 1/2
− (α2 B ′ − 2αβB ′ γ(3, 2, B ′ ) − (α2 A′ − 2αβA′ ) γ(3/2, A′ )
 

3 
− αβ γ(5/2, B) − γ(5/2, A) − γ(5/2, B ′ ) + γ(5/2, A′ )

5
3 ′ ′
− αβ e−B (1 + B) − e−A (1 + A) − e−B (1 + B ′ ) + e−A (1 + A′ ) .
 
2
(359)

Region III (E1 < Ef , E2 > Ef )

Z E2
G(E ′ , Ef ) dE ′
p
3 Ef Tm
E1
 2 1 2 1
= ( α2 B 5/2 − αβB 2 )E1 (B) − ( α2 A5/2 − αβA2 )E1 (A)

5 2 5 2
 2 5/2 1 2 2 5/2 1 2
− ( α2 B ′ + αβB ′ )E1 (B ′ ) − ( α2 A′ − αβA′ )E1 (A′ )

5 2 5 2
+ (α2 B − 2αβB 1/2 ) γ(3/2, B) − (α2 A − 2αβA1/2 ) γ(3, 2, A)
 

1/2 1/2
− (α2 B ′ + 2αβB ′ ) γ(3/2, B ′ ) − (α2 A′ − 2αβA′ ) γ(3/2, A′)
 

3 
− α2 γ(5/2, B) − γ(5/2, A) − γ(5/2, B ′ ) + γ(5/2, A′ )

5
3 ′ ′
− αβ e−B (1 + B) − e−A (1 + A) + e−B (1 + B ′ ) + e−A (1 + A′ ) − 2 .
 
2
(360)

8.15 File 6 Energy-Angle Distributions

If the File 6 data are expressed as a continuous energy-angle distribution (Law 1)
in the laboratory system, it is fairly easy to generate the multigroup transfer ma-
trix. As usual for GROUPR, the task is to calculate the “feed function” (the Leg-
endre moments for transferring to all possible secondary-energy groups starting
with incident energy E). The E ′ integration is controlled by the getmf6 sub-
routine, which calls f6lab to generate the integrands. This routine expects data
in Law-1 format, where the looping order is E, E ′ , µ. The only problems here
are handling the new ENDF-6 interpolation laws for two-dimensional tabulations
(for example, unit base) and converting tabulated angle functions for E→E ′ into
Legendre coefficients (which is done with a Gauss-Legendre quadrature of order
∼ 8).
If the File 6 data are given in the energy-angle form of Law 7 (where the

NJOY2016 223
LA-UR-17-20093 8 GROUPR

looping order is E, µ, E ′ ), GROUPR converts it to the Law 1 form using sub-

routine ll2lab. It does this by stepping through an E ′ grid that is the union of
the E ′ grids for all the different angles given. At each of these union E ′ values, it
calculates the Legendre coefficients using a Gauss-Legendre quadrature of order
8, and it checks back to see if the preceding point is still needed to represent the
distribution to the desired degree of accuracy. Now getmf6 and f6lab can be
used to complete the calculation as above.
If the File 6 data are expressed in the CM system, or if the phase-space
option is used, more processing is necessary to convert to the LAB system. This
conversion is done for each incident energy E given in the file. The grid for
laboratory secondary energy EL′ is obtained by doing an adaptive reconstruction
of the emission probability pLℓ (E, EL′ ) such that each Legendre order can be
expressed as a linear-linear function of EL′ . This part is done in subroutine
cm2lab. The values for pLℓ (E, EL′ ) are obtained in subroutine f6cm by doing an
adaptive integration along the contour EL′ =const in the EC′ , µC plane using µL
as the variable of integration:
Z +1
pLℓ (E, EL′ ) = pC (E, EC′ , µC ) Pℓ (µL ) J dµL , (361)
µmin

where µ is a scattering cosine and L and C denote the laboratory and center-of-
mass (CM) systems, respectively. The Jacobian is given by
s
EL′ 1
J= =p , (362)
EC′ 2
1 + c − 2cµL
and the cosine transformation is given by

µC = J(µL − c) . (363)

The constant c is given by

s s
A′ E
c= , (364)
(A + 1)2 EL′

where A is the ratio of the atomic weight of the target to the atomic weight of
the projectile, and A′ is the ratio of the atomic weight of the emitted particle
to the atomic weight of the projectile. The lower limit of the integral depends
on the maximum possible value for the center-of-mass (CM) secondary energy
as follows:

224 NJOY2016
8 GROUPR LA-UR-17-20093

1 E′ 
µmin = 1 + c2 − Cmax , (365)
2c EL′
where
r s
′
′
 ECmax A ′ 2
ELmax =E + . (366)
E (A + 1)2

An example of the integration contours for this coordinate transformation is

given in Fig. 30.
1.5

EL/E=1.0
1.0
EC/E

EL/E=0.5
0.5

EL/E=0.3

EL/E=.07
0.0

-1.0 -0.5 0.0 0.5 1.0

µC

Figure 30: Coordinate mapping between CM and laboratory reference frames for A=2. The
parameters EL and EC are the secondary energies in the Lab and CM frames.
The crosses on the curves are at µL values of -1.0, -.75, -.50, . . . , .75, and 1.0.
This figure also illustrates the advantage of integrating over µL for the contour
in the lower-left corner. The values of EC and µC are single-values functions of
µL .

The CM energy-angle distribution can be given as a set of Legendre coef-

ficients or a tabulated angular distribution for each possible energy transfer
E→E ′ , as a “precompound fraction” r(E, E ′ ) for use with the Kalbach-Mann[59]
or Kalbach[60] angular distributions, or as parameters for a phase-space distribu-
tion. The first three options are processed using f6ddx, and the last using f6psp.
The Kalbach option leads to a very compact representation. Kalbach and Mann

NJOY2016 225
LA-UR-17-20093 8 GROUPR

examined large number of experimental angular distributions for neutrons and

charged particles. They noticed that each distribution could be divided into two
parts: an equilibrium part symmetric in µ, and a forward-peaked preequilib-
rium part. The relative amount of the two parts depended on a parameter r,
the preequilibrium fraction, that varies from zero for low E ′ to 1.0 for large E ′ .
The shapes of the two parts of the distributions depended most directly on E ′ .
This representation is very useful for preequilibrium statistical-model codes like
GNASH[61], because they can compute the parameter r, and all the rest of the
angular information comes from simple universal functions. More specifically,
Kalbach’s latest work says that

a h i
f (µ) = cosh(aµ) + r sinh(aµ) , (367)
2 sinh(a)
where a is a simple function of E, E ′ , and Bb , the separation energy of the
emitted particle from the liquid-drop model without pairing and shell terms. The
separation energies are computed by formulas in bach. There is a problem for
elemental evaluations, because the calculations needs an A value for the element,
and it is difficult to guess which A value is most characteristic of the element. A
short table is included in the routine, and an “error in bach” will result if the
function is called for an element that doesn’t appear in the table. Similar routines
appear in HEATR and ACER. A better long-range solution would be desirable.
Fortunately, elemental evaluations are rare in modern evaluated libraries.
The File 6 processing methods in GROUPR apply equally well to neutrons,
photonuclear photons, and charged particles. The effects on the kinematics due
to the difference in mass between the incident particle and the emitted particle
are handled by the variable A′ in the above equations.

8.16 Smoothing
For continuum CM distributions in File 6, the low-energy shape should go like
EC . From Eq. 362, we see that the low-energy shape for the Lab distribution
p ′

would then go like EL′ . However, ENDF/B-VII evaluations for File 6 were
p

normally prepared using advanced nuclear model codes, such as GNASH from
LANL. These codes naturally produce spectra represented with histogram bins,
and these bins normally give a constant probability from zero energy to the
√
next bin boundary. This is certainly not a E shape! The histogram shape
can cause trouble with the CM-Lab transformation if some care is not taken.
Clearly, the histogram grossly overestimates the probability of scattering to very

226 NJOY2016
8 GROUPR LA-UR-17-20093

low energies. This apparent problem is somewhat alleviated by the fact that the
total probability of that lowest histogram bin is usually quite small. However,
in order to get better looking shapes, GROUPR has coding that replaces the
√
coarse histogram with a finer histogram chosen to represent a E shape more
closely. This smoothing coding scans up through the histogram distribution from
√
the evaluation to find the region that behaves like E. It then uses a recursive
procedure to subdivide the region using a given factor until it reaches a fairly
low energy (currently 40 eV).
√
A similar procedure is used to provide a E shape at low energies for delayed
neutron spectra.
For a few of the ENDF/B-VII actinides, the energy grid used to represent the
fission spectrum becomes too coarse above 10 MeV. If that situtation is found,
GROUPR changes the interpolation law from the given lin-lin option to lin-log
— that is, an exponential tail is assumed for the high energies. This assumption
is consistent with the orginal evaluations. This modification can be important
for reaction rates of high-threshold reactions.
These smoothing options are controlled by the global parameter ismooth,
which is turned on by default (in constrast to NJOY99 where it is turned off by
default).

8.17 GENDF Output

The group constants produced by GROUPR are normally written to an output
file in GENDF (groupwise ENDF) format for use by other modules of NJOY.
For example, DTFR can be used to convert a GENDF material to DTF (or
“transport”) format; CCCCR produces the standard interface files[11] ISOTXS,
BRKOXS, and DLAYXS; MATXSR produces a file in MATXS format[12]; and
POWR produces libraries for the Electric Power Research Institute (EPRI) codes
EPRI-CELL and EPRI-CPM. Other formats can easily be produced by new
modules, and some functions such as group collapse, are conveniently performed
directly in GENDF format. The GENDF format is also used for GAMINR
output, and both ACER and ERRORR use GENDF input for some purposes.
Depending on the sign of ngout2 (see below), the GENDF file will be written
in either coded mode (e.g., ASCII) or in the special NJOY blocked-binary mode.
Conversion between these modes can be performed subsequently by the MODER
module.
The GENDF material begins with a header record (MF=1, MT=451), but
the format of this first section is different from MT=451 on an ENDF or PENDF

NJOY2016 227
LA-UR-17-20093 8 GROUPR

tape. The section consists of a “CONT” record, containing

ZA Standard 1000*Z+A value

AWR Atomic weight ratio to neutron
0 Zero
NZ Number of σ0 values
-1 Identifies a GENDF-type data file
NTW Number of words in title
and a single “LIST” record, containing

TEMPIN Material temperature (Kelvin)

0. Zero
NGN Number of neutron groups
NGG Number of photon groups
NW Number of words in LIST
0 Zero
TITLE Title from GROUPR input (NTW words)
SIGZ Sigma-zero values (NZ words)
EGN Neutron group boundaries, low to high (NGN+1 words)
EGG Photon group boundaries, low to high (NGG+1 words).
For photoatomic GENDF files produced by the GAMINR module, the photon
group structure is stored in ngn and egn, and the number of photon groups is
given as ngg=0. The word count is NW=NTW+NZ+NGN+1+NGG+1. The
LIST record is followed by a standard ENDF file-end record (FEND). The normal
ENDF section-end (SEND) is omitted.
This header is followed by a series of records for reactions. The ENDF order-
ing requirements are relaxed, and MF and MT values can occur in any order.
Each section starts with a “CONT” record.
ZA Standard 1000*Z+A value
AWR Standard atomic weight ratio
NL Number of Legendre components
NZ Number of sigma-zero values
LRFLAG Break-up identifier flag
NGN Number of groups
It is followed by a series of LIST records, one for every incident-energy group
with nonzero result,

228 NJOY2016
8 GROUPR LA-UR-17-20093

TEMP Material temperature (Kelvin)

0. Zero
NG2 Number of secondary positions
IG2LO Index to lowest nonzero group
NW Number of words in LIST
IG Group index for this record
A(NW) Data for LIST (NW words),
where NW=NL*NZ*NG2. The last LIST record in the sequence is the one with
IG=NGN. It must be given even if its contents are zero. The last LIST record is
followed by a SEND record.
The contents of A(NW) change for various types of data. For simple cross
section “vectors” (MF=3), NG2 is 2, and A contains the two Fortran arrays
FLUX(NL,NZ), SIGMA(NL,NZ)
in that order. For ratio quantities like fission ν, NG2 is 3, and A contains
FLUX(NL,NZ), RATIO(NL,NZ), SIGMA(NL,NZ).
For transfer matrices (MF=6, 16, 21, etc.), A contains
FLUX(NL,NZ), MATRIX(NL,NZ,NG2-1).
The actual secondary group indices for the last index of MATRIX are usually
IG2LO, IG2LO+1, etc., using the GROUPR convention of labeling groups in order
of increasing energy. If the low-energy part of the fission matrix (or the fission
or capture photon production matrices) uses the special format described in
Section 8.6, the spectrum will be found in a LIST record with IG=0 and the
production cross section will be found in a series of records with IG2LO=0. The
group range for the spectrum ranges from IG2LO to IG2LO+NG2-1. For IG2LO=0,
NG2 will be 2 as for a normal cross section, and the two values will be the flux
for group IG and the corresponding production cross section.
Finally, for delayed neutron spectra (MF=5), NL is used to index the time
groups, NZ is 1, and there is only one incident energy record (IG=IGN). The
array A contains
LAMBDA(NL), CHID(NL,NG2-1),
where LAMBDA contains the delayed-neutron time constants and CHID contains
the spectra.
The GENDF material ends with a material-end (MEND) record, and the
GENDF tape ends with a tape-end (TEND) record.

NJOY2016 229
LA-UR-17-20093 8 GROUPR

8.18 Running GROUPR

GROUPR’s input instructions follow. They are reproduced from the comment
cards at the beginning of the 2016.0 version of the GROUPR module. Because
the code changes from time to time, it is a good idea to check these comment
cards in the current version to obtain up-to-date input instructions.

!---input specifications (free format)---------------------------

!
! card1
! nendf unit for endf tape
! npend unit for pendf tape
! ngout1 unit for input gout tape (default=0)
! ngout2 unit for output gout tape (default=0)
! card2
! matb material to be processed
! if ngout=0, matb<0 is an option to automatically
! process all the mats on the endf input tape.
! otherwise, matb<0 is a flag to add mts to and/or
! replace individual mts on gout1.
! ign neutron group structure option
! igg gamma group structure option
! iwt weight function option
! lord legendre order
! ntemp number of temperatures (default=1)
! nsigz number of sigma zeroes (default=1)
! iprint long print option (0/1=minimum/maximum)
! (default=1)
! ismooth switch on/off smoothing operation (1/0, default=1=on)
! set ismooth to 1 to enable sqrt(e) smoothing for
! mf6 cm emission spectra at low energies and for
! histogram delayed neutron spectra at low energies.
! card3
! title run label (up to 80 characters delimited by quotes,
! ended with /) (default=blank)
! card4
! temp temperatures in kelvin
! card5
! sigz sigma zero values (including infinity)
!
! if ign=1, read neutron group structure (6a and 6b)

230 NJOY2016
8 GROUPR LA-UR-17-20093

! card6a
! ngn number of groups
! card6b
! egn ngn+1 group breaks (ev)
!
! if igg=1, read gamma group structure (7a and 7b)
! card7a
! ngg number of groups
! card7b
! egg ngg+1 group breaks (ev)
!
! weight function options (8a,8b,8c,8d)
! card8a flux calculator parameters (iwt.lt.0 only)
! fehi break between computed flux and bondarenko flux
! (must be in the resolved resonance range)
! sigpot estimate of potential scattering cross section
! nflmax maximum number of computed flux points
! ninwt tape unit for new flux parameters (default=0)
! note: weighting flux file is always written binary
! jsigz index of reference sigma zero in sigz array
! (default=0)
! alpha2 alpha for admixed moderator (def=o=none)
! sam admixed moderator xsec in barns per absorber
! atom (def=0=none)
! beta heterogeneity parameter (def=0=none)
! alpha3 alpha for external moderator (def=0=none)
! gamma fraction of admixed moderator cross section in
! external moderator cross section (def=0)
! card8b tabulated (iwt=1 or -1 only)
! wght read weight function as tab1 record,
! this may span multiple lines and ends with a /.
! card8c analytic flux parameters (iwt=4 or -4 only)
! eb thermal break (ev)
! tb thermal temperature (ev)
! ec fission break (ev)
! tc fission temperature (ev)
! card8d input resonance flux (iwt=0 only)
! ninwt tape unit for flux parameters (binary)
!
! card9
! mfd file to be processed
! mtd section to be processed

NJOY2016 231
LA-UR-17-20093 8 GROUPR

! mtname description of section to be processed

! repeat for all reactions desired
! mfd=0/ terminates this temperature/material.
! card10
! matd next mat number to be processed
! matd=0/ terminates groupr run.
!
!---options for input variables----------------------------------
!
! ign meaning
! --- -------
! 1 arbitrary structure (read in)
! 2 csewg 239-group structure
! 3 lanl 30-group structure
! 4 anl 27-group structure
! 5 rrd 50-group structure
! 6 gam-i 68-group structure
! 7 gam-ii 100-group structure
! 8 laser-thermos 35-group structure
! 9 epri-cpm 69-group structure
! 10 lanl 187-group structure
! 11 lanl 70-group structure
! 12 sand-ii 620-group structure
! 13 lanl 80-group structure
! 14 eurlib 100-group structure
! 15 sand-iia 640-group structure
! 16 vitamin-e 174-group structure
! 17 vitamin-j 175-group structure
! 18 xmas nea-lanl
! all new additional group structure with 7 significant
! decimal digits compatible with calendf
! 19 ecco 33-group structure
! 20 ecco 1968-group structure
! 21 tripoli 315-group structure
! 22 xmas lwpc 172-group structure
! 23 vit-j lwpc 175-group structure
! 24 shem cea 281-group structure
! 25 shem epm 295-group structure
! 26 shem cea/epm 361-group structure
! 27 shem epm 315-group structure
! 28 rahab aecl 89-group structure
! 29 ccfe 660-group structure (30 MeV)

232 NJOY2016
8 GROUPR LA-UR-17-20093

! 30 ukaea 1025-group structure (30 MeV)

! 31 ukaea 1067-group structure (200 MeV)
! 32 ukaea 1102-group structure (1 GeV)
! 33 ukaea 142-group structure (200 MeV)
! 34 lanl 618-group structure
!
! igg meaning
! --- -------
! 0 none
! 1 arbitrary structure (read in)
! 2 csewg 94-group structure
! 3 lanl 12-group structure
! 4 steiner 21-group gamma-ray structure
! 5 straker 22-group structure
! 6 lanl 48-group structure
! 7 lanl 24-group structure
! 8 vitamin-c 36-group structure
! 9 vitamin-e 38-group structure
! 10 vitamin-j 42-group structure
!
! iwt meaning
! --- -------
! 1 read in smooth weight function
! 2 constant
! 3 1/e
! 4 1/e + fission spectrum + thermal maxwellian
! 5 epri-cell lwr
! 6 (thermal) -- (1/e) -- (fission + fusion)
! 7 same with t-dep thermal part
! 8 thermal--1/e--fast reactor--fission + fusion
! 9 claw weight function
! 10 claw with t-dependent thermal part
! 11 vitamin-e weight function (ornl-5505)
! 12 vit-e with t-dep thermal part
! -n compute flux with weight n
! 0 read in resonance flux from ninwt
!
! mfd meaning
! --- -------
! 3 cross section or yield vector
! 5 fission chi by short-cut method
! 6 neutron-neutron matrix (mf4/5)

NJOY2016 233
LA-UR-17-20093 8 GROUPR

! 8 neutron-neutron matrix (mf6)

! 12 photon prod. xsec (photon yields given, mf12)
! 13 photon prod. xsec (photon xsecs given, mf13)
! 16 neutron-gamma matrix (photon yields given)
! 17 neutron-gamma matrix (photon xsecs given)
! 18 neutron-gamma matrix (mf6)
! note: if necessary, mfd=13 will automatically change
! to 12 and mfd=16 will automatically change to 17 or 18.
! 21 proton production matrix (mf6)
! 22 deuteron production (mf6)
! 23 triton production (mf6)
! 24 he-3 production (mf6)
! 25 alpha production (mf6)
! 26 residual nucleus (a>4) production (mf6)
! 31 proton production matrix (mf4)
! 32 deuteron production (mf4)
! 33 triton production (mf4)
! 34 he-3 production (mf4)
! 35 alpha production (mf4)
! 36 residual nucleus (a>4) production (mf4)
! note: if necessary, mfd=21-26 will
! automatically change to 31-36.
! 1zzzaaam nuclide production for zzzaaam
! subsection from file 3
! 2zzzaaam nuclide production for zzzaaam
! subsection from file 6
! 3zzzaaam nuclide production for zzzaaam
! subsection from file 9
! 4zzzaaam nuclide production for zzzaaam
! subsection from file 10
! 40000000 fission product production (mtd=18 only)
! subsection from file 10
!
! mtd meaning
! --- -------
! -n process all mt numbers from the previous
! entry to n inclusive
! 221-250 reserved for thermal scattering
! 257 average energy
! 258 average lethargy
! 259 average inverse velocity (m/sec)
!

234 NJOY2016
8 GROUPR LA-UR-17-20093

! automatic reaction processing options

! -------------------------------------
! 3/ do all reactions in file3 on input pendf
! 6/ do all matrix reactions in endf dictionary
! 10/ do all isotope productions using mf8
! 13/ do all photon production cross sections
! 16/ do all photon production matrices
! 21/ do all proton production matrices
! 22/ do all deuteron production matrices
! 23/ do all triton production matrices
! 24/ do all he-3 production matrices
! 25/ do all alpha production matrices
! 26/ do all a>4 production matrices
!
!-------------------------------------------------------------------

In these instructions, card1 defines the input and output units for GROUPR.
The module requires both ENDF and PENDF input tapes, because the PENDF
tapes produced by RECONR, BROADR etc., do not contain angle (MF=4),
energy (MF=5), or photon (MF=12, 15) distributions. For materials that do
not use resonance parameters to represent part of the cross section, it is possible
to use a copy of the ENDF tape in place of the PENDF tape. The normal mode
for GROUPR is to use ngout1=0; however, sometimes it is convenient to add a
new material or reaction to an existing GENDF tape. The old GENDF tape is
then mounted on unit ngout1, and the revised GENDF tape will be written to
ngout2.
Card 2 selects the first material to be processed (matb) and sets up the group
structures, weighting option, Legendre order, and self-shielding parameters for
all the materials to be processed in this run.
The names of the available group structures are given in the input instruc-
tions. Energy bounds or lethargy bounds can be found in the source code. Of
course, it is always possible to read in an arbitrary group structure (see card6a
through card7b). The energies must be given in increasing order (note that this
is opposite from the usual convention). Here is an example of the input cards
for the conventional 4-group structure historically used in some thermal reactor
codes:

4/ card6a

NJOY2016 235
LA-UR-17-20093 8 GROUPR

Figure 31: Built-in neutron weight functions of GROUPR on a logarithmic flux-per-unit-

lethargy plot that emphasizes the low energy range.

1e-5 .625 5530 .821e6 10e6 / card6b

These cards are read by the standard Fortran READ* method. Fields are delim-
ited by space, and “/” terminates the processing of input on a card. Anything
after the slash is a comment.
The available weight function options are listed in the input instructions under
iwt. See Fig. 31 and Fig. 32. Here are brief descriptions of the options:

IWT=2 The weight function is constant (not shown in the Figures). This
option is usually chosen for very fine group structures such as the 620-group
or 640-group dosimetry structures.
IWT=3 The weight function is proportional to 1/E. The slowing-down of
neutrons in water gives a 1/E flux from about 1 eV up to 100 keV, or so.
This weight function is traditionally used for calculating resonance integrals,
but it is not useful at the lower and higher energies needed for a full set of
transport constants. Although not shown, the graph of this function is a
flat line on a flux-per-unit-lethargy plot, such as the one in Fig. 31.
IWT=4 This weight function combines a thermal Maxwellian at low energies,
a 1/E function at intermediate energies, and a fission spectrum at high
energies to obtain a function appropriate for several different applications.
The temperatures of the Maxwellian and fission parts and the energies

236 NJOY2016
8 GROUPR LA-UR-17-20093

Figure 32: Built-in neutron weight functions on a linear plot that emphasizes the high-energy
range.

where the components join can be chosen by the user. Therefore, option 4
can be used to produce typical thermal reactor weight functions like those
shown in Figures 31 and 32, a pure fission spectrum for calculating some
kinds of dosimetry cross sections, or a pure thermal spectrum for getting
effective thermal average cross sections. The function for IWT=4 shown
in the figure was produced using a thermal temperature of 0.025 eV joined
to 1/E at 0.1 eV, and a fission temperature of 1.40 MeV joined to 1/E at
820.3 keV.
IWT=5 This is a mid-life PWR (pressurized water reactor) flux spectrum
with a fusion peak added (see below for a discussion of the fusion peak
used). Note the peaks and dips resulting from oxygen resonances and win-
dows at high energies, and the hardening apparent in the epithermal region.
The thermal part of the function is also hardened with respect to a simple
Maxwellian shape. The dips that should show up in the eV range due to
resonances and 238 U have been removed to allow the self-shielding method
to work without risk of counting the shielding effects twice. This weight
function has been used for several libraries[62] prepared for the Electric
Power Research Institute (EPRI), and for the MATXS7 library used with
the TRANSX code.
IWT=6 This function is similar to option 4, but the breakpoints were chosen
to keep the curves continuous. The thermal Maxwellian is calculated for
300K. In this case, the fusion peak (see below) was added to the high-energy
tail of the fission spectrum for smoothness.
IWT=7 This option is reserved for future use.

NJOY2016 237
LA-UR-17-20093 8 GROUPR

IWT=8 This function is intended for cross sections libraries used for fast
reactor analysis (typically, fast breeder designs), but is also useful for fusion-
blanket problems. It has a fusion peak at high energies, followed by a fission
spectrum, a slowing-down spectrum typical of a fast reactor, and a thermal
tail. The tail is provided to help give reasonable results in shields far from
the core; its characteristic temperature is 300K. Note the sharp drop in
the flux as energy decreases from 19 keV. This region is important for 238 U
absorption, and this drop-off helps to give good group constants for fast
reactors. Of course, it would be entirely wrong for a thermal reactor.
IWT=9 This is another typical thermal+1/E+fission+fusion function that
has been used for many libraries at LANL in the 30-group structure. The
CLAW-3 and CLAW-4 libraries are available from the Radiation Shielding
Information Computational Center at ORNL.
IWT=10 This is the same as the CLAW weight function (IWT=9), but
the shape of the thermal part is automatically recalculated to follow a
Maxwellian law for temperature T .
IWT=11 This is the weight function used in the VITAMIN-E library. It has
the following segments: a .0253-eV thermal Maxwellian below 0.414 eV, a
1/E law from .414 eV to 2.12 MeV, a 1.415-MeV fission spectrum from 2.12
to 10 MeV, another 1/E section from 10 to 12.52 MeV, a fusion peak (25
keV width) between 12.52 and 15.68 MeV, and a final 1/E section for all
higher energies. The shape of the fusion peak is almost identical to IWT=5
(see Fig. 32). The low-energy part of this weight function is not shown.

Just as in the case of group structures, an arbitrary weight function can be read
in (see card8b). This function is presented to GROUPR as an ENDF/B “TAB1”
record. This means that a count of E, C(E) pairs and one or more interpolation
schemes are given.
An ENDF/B “TAB1” record consists of three distinct parts:
• two double values and four integer values of which only the last two integers
(the number of interpolation ranges NR and the number of E, C(E) pairs
NP) are needed to read the remainder of the record
• the interpolation scheme data which is a sequence of NR pairs of NBT (index
of the E, C(E) pair corresponding to the end of interpolation range) and
INT (the interpolation type)
• the tabulated data which is a sequence of NP E, C(E) pairs
The interpolation type INT specifies the interpolation law to be used in the
interpolation range. It can take the following values:

238 NJOY2016
8 GROUPR LA-UR-17-20093

INT Meaning
1 histogram
2 linear-linear
3 linear-log
4 log-linear
5 log-log

For example, INT=5 specifies that lnC is a linear function of lnE. Similarly,
INT=4 specifies that lnC is a linear function of E.
In its most general form, these input cards would be

0. 0. 0 0 NR NP
NBT(1) INT(1) ... NBT(NR) INT(NR)
E(1) C(1) ... E(NP) C(NP) / card8b

For example, a function using two interpolation ranges: the first one between
1e-5 eV and 100 eV using a histogram and the second one between 100 eV and
20 MeV using lin-lin interpolation)

0. 0. 0 0 2 6
3 1 6 2
1e-5 0.5 1. 0.75 100. 0.8
1e+6 0.85 1.5e+6 0.9 2.0e+7 1.0 / card8b

For the special case of a single interpolation scheme, the input cards are simplified
as follows

0. 0. 0 0 1 NP
NP INT
E(1) C(1) ... E(NP) C(NP) / card8b

As many physical lines as are needed can be used for “card8b”, as long as the
terminating slash is included.
One of the weighting options, IWT=4, is a generalized “1/E+fission+thermal”
function where the thermal temperature, fission temperature, and breakpoint en-

NJOY2016 239
LA-UR-17-20093 8 GROUPR

ergies (all in eV) are given on card8c. The weight function for the Los Alamos
LIB-IV cross section library[63] used

.10 .025 820.3e3 1.4e6 / card8c

See Figures 31 and 32 for a plot of this function.

Several of these weight functions include a fusion peak. Because of the finite
width of the distribution of ion energies in a D-T fusion plasma, the emitted
14-MeV neutrons clearly will not have a delta-function energy spectrum. In fact,
owing to the presence of a cross-product term in the kinematic relations, the
typical ion-energy spread of a few tens of kilovolts is magnified into a neutron-
energy spread of around 1 MeV. For an assumed isotropic Maxwellian plasma,
the neutron peak shape (for example, see the review article by Lehner[64]) is

∞
g 3 σ(g)
Z
exp{−b v 2 + v0 (g)2 − cg 2 } sinh(2bvv0 ) (368)
 
S(E) = C dg .
0 v0 (g)

Here S(E) is the number of neutrons emitted with laboratory energy between
E and E + dE, C is a normalization factor, v is the laboratory velocity corre-
sponding to energy E, and v0 is the velocity of the neutron in the CM system.
Both v0 and the fusion cross section σ are determined by the relative velocity
g of the reacting ions, the integration variable in Eq. 368. The coefficient b is
equal to M/2kT , where M is the total mass of the reacting ions and kT is the
plasma temperature. Similarly, c is µ/2kT , where µ is the reduced mass of the
ion system. The only approximation involved in Lehner’s derivation of Eq. 368 is
that all particles may be treated nonrelativistically. At 14 MeV, the relativistic
factor

1
γ=p , (369)
1 − v 2 /c2
is very close to unity (1.015), and it varies negligibly over the range of interest
(say, 13 to 17 MeV). It is sufficient then to invoke relativistic mechanics in
defining the location of the 14-MeV peak but not in discussing the shape. This
effect moves the peak toward lower neutron energies, but only by about 20 keV.
Although the expression for S(E) in Eq. 368 is accurate, it has the disadvantage
of requiring a numerical integration at each point E in the energy spectrum. For
this reason, we consider what simplifications can be made without serious loss

240 NJOY2016
8 GROUPR LA-UR-17-20093

of accuracy. In the energy range around the 14-MeV peak, the product 2bvv0
in Eq. 368 has a numerical value of about 6000. Thus, the hyperbolic sine can
obviously be replaced by just the positive exponential term. If we make this
change in Eq. 368, we can write the following, still nearly exact, expression for
the neutron spectrum:
Z ∞
S1 (E) = C1 exp[−b(v − v0 )2 ] P (v0 ) dv0 . (370)
0

Here we also have inverted the function v0 (g) and changed the integration vari-
able. The spectrum then is a linear superposition of velocity exponentials with
slightly different peak locations. For normal plasma temperatures, the velocity
distribution P (v0 ) is very narrow, since the expression for v0 is dominated by
the nuclear Q-value (17.586 MeV) rather than the contribution from the ion ki-
netic energy (typically around 50 keV). Thus, it seems reasonable to approximate
P (v0 ) as delta function

P (v0 ) ≈ δ(v0 − vp ) . (371)

This gives a second approximate form,

S2 (E) = C2 exp[−b(v − vp )2 ] , (372)

where vp has the obvious meaning of the laboratory neutron velocity at the center
of the peak. We shall refer to this as the velocity exponential form of the neutron
energy spectrum. An expression essentially identical to Eq. 372 was given in an
early paper by Nagle and coworkers[65]. In order to examine the accuracy of
the velocity exponential form, we have calculated S(E) from Eq. 368 and S2 (E)
from Eq. 372 at 20 keV, a typical plasma temperature in current fusion-reactor
concepts. In performing the numerical integration over g in Eq. 368, we used
numerical values for the D-T fusion cross section taken from the compilation
by Jarmie and Seagrave[66]. In evaluating S2 (E) using Eq. 372, a value of vp
was chosen so as to force agreement between S2 (E) and S(E) at 17 MeV. As
discussed by Muir[67], the overall agreement is remarkably good, the maximum
error over the range from 13.5 to 17 MeV being about 2%. The value of vp thus
derived corresponds to a peak-center energy Ep of 14.07 MeV. This value includes
the small (∼ 20 keV) relativistic correction mentioned above. If we approximate
the mass of the D+T system as 5 times the neutron mass, then we obtain the
recommended peak shape

NJOY2016 241
LA-UR-17-20093 8 GROUPR

 5 √
Srec (E) = exp − ( E − Ep )2 , (373)
p 
kT
where Ep =14.07 MeV. The functional form in Eq. 373 was used to calculate the
fusion peak shapes appearing in two of the data statements in genwtf (namely,
those utilized for IWT=5 and 8) and is also used explicitly in getwtf to calculate
analytically the weight function for IWT=6. In all three cases, kT =25 keV is
used as an average or typical fusion-reactor plasma temperature. See Fig. 32 for
a graphical display of the resulting weight functions in the 14-MeV region.
The GROUPR flux calculator is selected by a negative sign on iwt. The
additional card8a is then read. The calculator option used is determined by
the number of parameters given and their values. The parameters fehi and
nflmax are used to select the energy range for the flux calculation, and they also
determine the cost in time and storage. The actual value for sigpot is not very
critical – a number near 10 barns is typical for fissionable materials.
Nonzero values for ninwt and jsigz will cause the computed flux for a given
fissionable isotope (such as 238 U) to be written out onto a file. This saved flux can
be used as input for a subsequent run for a fissile material (such as 239 Pu) with
iwt=0 to get an approximate correction for resonance-resonance interference.
See Eq. 278.
Nonzero values for some of the last five parameters on card8a select the
extended flux calculation of Eq. 276. The simplest such calculation is for an iso-
lated pin containing a heavy absorber with an admixed moderator. For 238 UO ,
2
the card might be

400 10.6 5000 0 0 .7768 7.5 / card8a

where 7.5 barns is twice the oxygen cross section and α is computed from (A −

2
1)/(A + 1) with A = 15.858. A more general case would be a PWR-like lattice
of 238 UO
2 fuel rods in water:

400 10.6 5000 0 0 .7768 7.5 .40 1.7e-7 0.086 / card8a

where 0.086 is computed using 3.75 barns for O and 40 barns for the two H
atoms bound in water; that is,

242 NJOY2016
8 GROUPR LA-UR-17-20093

3.75
γ= = 0.086 . (374)
2 ∗ 20 + 3.75

A third example would compute the flux for a homogeneous mixture of 238 U and
hydrogen

400 10.6 5000 0 0 0 0 1. 1.7e-7 / card8a

As a final example, consider a homogeneous mixture of uranium and water. This

requires beta=1 and sam=0. Thus,

400 10.6 5000 0 0 .7768 0. 1. 1.7e-7 .086 / card8a

The maximum Legendre expansion order used for scattering matrices is set
by lord. The number of tables produced is lord+1; that is, ℓ = 0, 1, ... lord.
When more than 1 value of σ0 is requested, both the ℓ=0 and ℓ=1 components
of the total cross section are produced.
Card 3 contains a short descriptive title that is printed on the listing and
added to the output GENDF tape. Card 4 gives the ntemp values of temperature
for the run. They must be in ascending order, and if unresolved data are included
on the PENDF tape, the temperatures in this list must match the first ntemp
values in MF=2, MT=152 from UNRESR or PURR (see stounr and getunr).
Card 5 gives the σ0 values for the run in descending order, starting with infinity
(represented by 1010 barns).
This completes the description of the global input parameters for GROUPR.
The rest of the input cards request reactions to be processed for the various
temperatures and materials desired. Because of the many types of data that it
can process, GROUPR does not have a completely automatic mode for choosing
reactions to be processed. On the basic level, it asks the user to request each
separate cross section or group-to-group matrix using the parameters mfd, mtd,
and mtname. However, simplified input modes are also available. For example,
the one “card9” containing

3/

NJOY2016 243
LA-UR-17-20093 8 GROUPR

will process the cross section “vectors” for all of the reaction MT numbers found
on the PENDF tape.
For completeness, the full input for matd, mfd, and mtname will be described
first. Most readers can skip to the description of automated processing below.
The value of mfd depends on the output desired (vector, matrix) and the form
of the data on the ENDF evaluation. Simple cross section “vectors” σxg are
requested using mfd=3 and the mtd numbers desired from the list of reactions
available in the evaluation (check the directory in MF=1,MT=451 of the ENDF
and PENDF tapes for the reactions available). A typical example would be

3 1 ’total’/
3 2 ’elastic’/
3 16 ’(n,2n)’/
3 51 ’(n,nprime)first’/
3 -66 ’(n,nprime)next’/
3 91 ’(n,nprime)continuum’/
3 102 ’radiative capture’/

The combinations of “3 51” followed by “3 −66” means process all the reac-
tions from 51 through 66; that is, (n,n′1 ), (n,n′2 ), . . . , (n,n′16 ). If self-shielding is
requested, the following reactions will be processed using nsigz values of back-
ground cross section: total (MT=1), elastic (MT=2), fission (MT=18 and 19),
radiative capture (MT=102), heat production (MT=301), kinematic KERMA
(MT=443), and damage energy production (MT=444). The other File-3 reac-
tions will be computed at σ0 =∞ only. This list of reactions can be altered by
small changes in init if desired.
There are several special options for mtd available when processing cross sec-
tion vectors:

mtd Option
259 Average inverse neutron velocity for group in s/m.
258 Average lethargy for group.
251 Average elastic scattering cosine µ computed from File 4.
252 Continuous-slowing-down parameter ξ (average logarithmic en-
ergy decrement for elastic scattering) computed from File 4.

244 NJOY2016
8 GROUPR LA-UR-17-20093

253 Continuous-slowing-down parameter γ (the average of the square

of the logarithmic energy decrement for elastic scattering, divided
by twice the average logarithmic energy decrement for elastic
scattering) computed from File 4.
452 ν: the average total fission yield computed from MF=1 and
MF=3.
455 ν D : the average delayed neutron yield computed from MF=1 and
MF=3.
456 ν P : the average prompt fission neutron yield computed from
MF=1 and MF=3.

There are also some special options for mfd that can be used when processing
cross sections:

mfd Option
12 Photon production cross section computed from File 12 and File
3.
13 Photon production cross section computed from File 13. Recent
versions of GROUPR will automatically shift between 12 and 13,
if necessary.
1zzzaaam nuclide production for zzzaaam from a subsection of MF=3
2zzzaaam nuclide production for zzzaaam from a subsection of MF=6
3zzzaaam nuclide production for zzzaaam from a subsection of MF=9
4zzzaaam nuclide production for zzzaaam from a subsection of MF=10
40000000 fission product production from the MT=18 subsection of MF=10

An example of the isomer production capability would be the radiative capture

reaction of ENDF/B-V 109 Ag(n,γ) from Tape 532:

30471090 102 ’(n,g) TO g.s.’/

30471091 102 ’(n,n) to isomer’/

Starting with ENDF/B-VIII.0, some non-fissile nuclides can have an MT=18

section (fission) in MF=10 (radioactive isotope production) to represent breakup
due to high energy particles. In these cases, it is often not possible to designate
a specific nuclide, which is why the mfd value is set to 40000000. In such a case,
the following input will make GROUPR process this part of the ENDF file:

NJOY2016 245
LA-UR-17-20093 8 GROUPR

40000000 18 ’HE breakup’ /

The next class of reactions usually processed is the group-to-group neutron

scattering matrices. The complete list of mtd values is most easily found under
File 4 in the MF=1,MT=451 “dictionary” section of the evaluation. An example
follows:

6 2 ’elastic matrix’/
6 16 ’(n,2n) matrix’/
6 51 ’(n,nprime)first matrix’/
6 -66 ’(n,nprime)next matrix’/
6 91 ’(n,nprime)continuum matrix’/ .

Using mfd=6 implies that File 4, or File 4 and File 5, will be used to generate the
group-to-group matrix. The elastic matrix will be computed for nsigz values of
background cross section, but the other reactions will be computed for σ0 =∞
only. The list of matrices to be self-shielded can be altered by changing init.
Fission is more complex. For the minor isotopes, only the total fission reaction
is used, and the following input is appropriate for the prompt component:

3 18 ’fission xsec’/
6 18 ’prompt fission matrix’/

For the important isotopes, partial fission reactions are given. They are really
not needed for most fission reactor problems, and the input above is adequate.
However, for problems where high-energy neutrons are important, the following
input should be used:

3 18 ’total fission’/
3 19 ’(n,f)’/
3 20 ’(n,nf)’/
3 21 ’(n,2nf)’/
3 38 ’(n,3nf)’/
6 19 ’(n,f)’/

246 NJOY2016
8 GROUPR LA-UR-17-20093

6 20 ’(n,nf)’/
6 21 ’(n,2nf)’/
6 38 ’(n,3nf)’/

Note that “6 18” is omitted because it will, in general, be different from the sum
of the partial matrices (see Section 8.6). Some materials don’t have data for
(n,3nf); in these cases, omit the two lines with mtd=38 from the input. The
fission matrix is not self-shielded. Since resonance-to-resonance fission-spectrum
variations are not described in the ENDF format, it is sufficient to self-shield
the cross section and then to use the self-shielding factor for the cross section to
self-shield the fission neutron production.
Delayed fission data are available for the important actinide isotopes, and the
following input to GROUPR is used to process them:

3 455 ’delayed nubar’/

5 455 ’delayed spectra’/

The line for mfd=5 automatically requests spectra for all time groups of de-
layed neutrons. The time constants are also extracted from the evaluation. As
discussed in Section 8.6, formatting modules such as DTFR and CCCCR must
combine the prompt and delayed fission data written onto the GENDF tape in
order to obtain steady-state fission parameters for use in transport codes.
Starting with the ENDF-6 format, neutron production data may also be found
in File 6, and mfd=8 is used to tell the code to use MF6 for this mtd. When
using full input, the user will have to check the File 1 directory and determine
what subsections occur in File 6.
Photon production reactions can be found in the ENDF dictionary under
MF=12 and 13. To request a neutron-to-photon matrix, add 4 to this number.8
For example,

17 3 ’nonelastic photons’/
16 4 ’inelastic photons’/
16 18 ’fission photons’/
8
In recent versions of NJOY, GROUPR will automatically shift between 16 and 17 using data read from
the ENDF dictionary by the conver subroutine. Thus, use of mfd=17 is no longer necessary.

NJOY2016 247
LA-UR-17-20093 8 GROUPR

16 102 ’capture photons’/

Yields (MF=12) are normally used with resonance reactions (MT=18 or MT=102),
or for low-lying inelastic levels (MT=51, 52, ....). MT=3 is often used by eval-
uators as a catch-all reaction at high energies where it is difficult to separate
the source reactions in total photon emission measurements. In these cases,
photon production cross sections from other reactions like MT=102 are nor-
mally set equal to zero at high energies. The general rule for photon emission
is that the total production is equal to the sum of all the partial production
reactions given in the evaluation. Starting with the ENDF-6 format, photon
production may also appear in File 6. Use mfd=18 to process these contribu-
tions. Since resonance-to-resonance variations in photon spectra are not given
in ENDF evaluations, GROUPR does not normally self-shield the photon pro-
duction matrices (although this can be done if desired by making a small change
in init); instead, it is assumed that only the corresponding cross section needs
to be shielded. Subsequent codes can use the cross section self-shielding fac-
tor with the infinite-dilution photon production matrix to obtain self-shielded
photon production numbers.
This version of GROUPR can also generate group-to-group matrices for charged-
particle production from neutron reactions and for all kinds of matrices for in-
cident charged particles. The incident particle is determined by the input tape
mounted. The identity of the secondary particle is chosen by using one of the
following special mfd values:

For distributions given in File 6 (energy-angle):

mfd Meaning
21 proton production
22 deuteron production
23 triton production
24 3 He production
25 alpha production
26 residual nucleus (A>4) production

For distributions given in File 4 (angle only):

248 NJOY2016
8 GROUPR LA-UR-17-20093

mfd Meaning
31 proton production
32 deuteron production
33 triton production
34 3 He production
35 alpha production
36 residual nucleus (A>4) production

If necessary, mfd=21-26 will automatically change to 31-36.

The user will normally process all reactions of interest at the first temperature
(for example, 300K). At higher temperatures, the threshold reactions should be
omitted, because their cross sections do not change significantly with tempera-
ture except at the most extreme conditions. This means that only the follow-
ing reactions should be included for the higher temperatures (if present): total
(MT=1), elastic (MT=2), fission (MT=18), radiative capture (MT=102), heat-
ing (MT=301), kinematic KERMA (MT=443), damage (MT=444), and any
thermal cross sections (MT=221-250). Only the elastic and thermal matrices
should be included at the higher temperatures.
Warning: when using the explicit-input option, it is a fatal error to request
a reaction that does not appear in the evaluation, cannot be computed from
the evaluation, or was not added to the PENDF tape by a previous module.
Reactions with thresholds above the upper boundary of the highest energy group
will be skipped after printing a message on the output file.
Automated processing of essentially all reactions included in an ENDF/B
evaluation is also available. As mentioned previously, the single card

3/

will process all the reactions found in File 3 of the input PENDF tape. How-
ever, this list excludes thermal data (MT=221-250) and special options such as
mtd=251-253, 258-259, and 452-456. If any of these reactions are needed, they
should be given explicitly (see example below). Similarly, the single card

6/

NJOY2016 249
LA-UR-17-20093 8 GROUPR

will process the group-to-group matrices for all reactions appearing in File 4
of the ENDF/B tape, except for MT=103-107 and thermal scattering matrices
(MT=221-250). If MT=18 and 19 are both present, only MT=19 will be pro-
cessed into a fission matrix. For ENDF-6 evaluations, the “8/” option will also
process every neutron-producing subsection in File 6. Photon production cross
sections are requested using

13/

and photon-production matrices are requested with the single card

16/

In both cases, all reactions in both File 12 and File 13 will be processed without
the need for using mfd=12 or mfd=17. For ENDF-6 libraries, this option will
also process all photon-production subsections in File 6. There is no automatic
option for delayed neutron data. An example of a processing run for a fissionable
isotope with thermal cross sections follows:

3/
3 221/ thermal xsec
3 229/ average inverse velocity
3 455/ delayed nubar
5 455/ delayed spectra
6/
6 221/ thermal matrix
16/ photon production matrix

An example of charged-particle processing for the incident-neutron part of a

coupled n-p-γ library follows:

3/ cross sections
6/ neutron production matrix
16/ photon production matrix

250 NJOY2016
8 GROUPR LA-UR-17-20093

21/ proton production matrix

The layout of data in a n-p-γ coupled set is shown in Figure 33.

edits

3 σa,νσf,σt

0
22
γ→ n
0

52
0

p→ n
γ→ p

52
γ→ γ, up
n→n, up p→p, up

n→ n, down p→ p, down
max likely γ→ γ, down
upscatter
n→ p
p→ γ

0
n→ γ

n (52) p (52) γ (22)

Figure 33: Layout of a coupled table for the simultaneous transport of neutrons, protons,
and gamma rays. Normally, only the lower triangle of the “p to n” block would
contain values for the upscatter portion of the table (the part above the line in
the middle). The “group” index increases from left to right, and the “position”
index increases from top to bottom.

There is a new ENDF format now becoming available that helps to describe
the production of all isotopes and isomers in a given system. It uses a directory
in File 8 to direct the code to productions represented by MF=3, by MF=9 times

NJOY2016 251
LA-UR-17-20093 8 GROUPR

MF=3, by MF=10, or by sections of MF=6 times MF=3. When this format is

used, it is possible to give the simple command

10/

to have production cross sections generated for every nuclide that is produced.
These production sections are labeled with the ZA and isomeric state of the
products for use by subsequent NJOY modules.

8.19 Coding Details

The groupr subroutine is exported by the groupm module. GROUPR begins by
reading most of the user’s input (see ruinb). It then locates the desired material
and temperature on the input ENDF, PENDF, and GENDF tapes, and reads in
the self-shielded unresolved cross sections (if any) from the PENDF tape using
stounr. If self-shielding was requested, genflx is used to compute the weighting
flux as described in Section 8.4. The next step is to write the header record for
this material on the output GENDF tape.
The code is now ready to begin the loop over reactions for this material
and temperature. Either an input card is read to get mfd, mtd and mtname
(the reaction name), or the next reaction in an automatic sequence is selected.
First, the default Legendre order, secondary group count, and σ0 count are
selected for the reaction in init, and the retrieval routines are initialized. groupr
then processes the reaction using the panel logic described in Section 8.11. If
a “shortcut” fission spectrum was requested (mfd=5), for delayed fission, and
for the low-energy “constant” spectra, the spectrum is calculated directly using
getff. As the cross sections for each group are obtained, they are printed out
(see displa) and written to the GENDF tape. When the last group has been
processed, groupr loops back to read a new input card for a new reaction.
This loop over reactions continues until a terminating “0/” card is read.
groupr then proceeds to the next temperature, if any, and repeats the loop
over reactions. After the last temperature has been processed for the first ma-
terial, an opportunity is provided to change to a new material, keeping all the
other input parameters unchanged. A “0/” card at this point causes all the files
to be closed, prints out the final messages, and terminates the groupr run.
Automatic choice of the next reaction to be processed is done in one of two
ways. For a simple range of MT numbers, such as the example 51 - 66 used above,

252 NJOY2016
8 GROUPR LA-UR-17-20093

the negative value is stored in the variable mtdp in the groupr subroutine. When
mtdp is negative, mtd is incremented after each reaction until it is greater than
the absolute value of mtdp. mtdp is then reset to one, and the code proceeds to
the next input card. When processing photon data, mfd=16 will automatically
change to 17, if necessary. Similarly, mfd=21, 22, . . . will automatically change
to mfd=31, 32, . . . . A more automatic method is triggered by mtd=0. In this
case, a subroutine called nextr is called to return the next value of mtd to be
used, and a subroutine called namer is called to generate the reaction name.
For mfd=3, nextr finds the next reaction in File 3 on the input PENDF tape.
MT=251-253 and thermal data (MT=221-250) are excluded. The MT values
for the special options (258, 259, etc.) do not appear on the PENDF tape, and
they must be requested explicitly. For matrices, GROUPR works with a set of
lists loaded into global arrays by conver. The list mf4 contains all the neutron-
scattering MT numbers that appear in the File 4 part of the directory on the
ENDF tape, and the list mf6 contains all the MT numbers of sections of File
6 that contain subsections that produce neutrons. Therefore, reading through
these two lists returns all the neutron-producing matrix reactions. Similarly, the
list mf12 contains all the File 12 entries from the directory, mf13 contains the File
13 entries, and mf18 contains all the MT values for sections in File 6 that contain
subsections for photon production. Scanning through these three lists produces
all the photon production matrix reactions. Two arrays are used for charged-
particle producing reactions; the first index runs through the charged particles
in the order p, d, t, 3 He, α, recoil. Taking proton production as an example,
the list elements mf6p(1,i) contain the MT numbers of sections in File 6 that
contain subsections that produce protons. The list elements mf4r(1,i) contain
MT numbers from File 4 for two-body reactions that produce protons; namely,
MT600 - MT648. nextr scans through both of these lists to return indexes to
all the reactions that produce protons. The same procedure is used for the other
charged particles. The arrays mf10s and mf10i are used in a similar way for
nuclide production.
Subroutine namer generates name strings with up to 15 Hollerith words with 4
characters each (60 characters). The names depend on the “ZA” of the projectile
and the MT number for the reaction. The parameter mfd is used to choose
between the suffixes “cross section” and “matrix”. Some examples of the names
produced follow:

NJOY2016 253
LA-UR-17-20093 8 GROUPR

Name Name Name

(n,total) (n,heat) (p,p02)
(n,elastic) (n,p02) (p,n00)
(n,2n) (p,elastic) (g,total)
(n,n01) (p,2n) (g,pair)

Subroutines mfchk and mfchk2 are used with the full input for reaction se-
lection in GROUPR to check whether mfd=17 is needed when mfd=16 was re-
quested, or whether one of the charged-particle File 4 values mfd=31-36 is needed
when mfd=21-26 were requested. The lists in global variables like mf12 are used,
just as for nextr.
Subroutine gengpn generates the group bounds in the global array egn for
the neutron group structure from input cards, from data statements, or by cal-
culation. Some of the data statements use energies in eV and some use lethargy.
Similarly, gengpg generates the photon group structure in global array egg from
input cards or data statements; in this case, all bounds are in eV.
Subroutine genwtf sets up the weight function option requested with iwt by
reading the input cards into weights, transferring numbers from data statements
to weights, or simply reporting the analytic weight option requested. Subroutine
getwtf returns the values of the weight function at energy E by calculation or
by interpolation in the table established by genwtf. The current version returns
the same value for all Legendre orders. Choosing enext is difficult for getwtf
because the functions have not been explicitly linearized. It is important to
generate extra grid points in energy regions where the weight function may vary
faster than the cross section (for example, in the fusion peak).
Subroutine genflx computes the self-shielded weighting flux using either the
Bondarenko model or the flux calculator, and it writes the result on a scratch
tape using the loada utility routine. When fluxes are needed for the generalized
group integrals, they are read from this scratch file using finda (see getflx).
Subroutine genflx starts by checking for the weighting option. If the Bon-
darenko model was selected, it initializes gety1 and getwtf to read the total
cross section from the PENDF tape and the smooth weighting function C(E)
set up by genwtf. It then steps through the union grid of σt (E) and C(E)
computing the flux vs. σ0 and Legendre order by means of Eq. 253. In the
unresolved energy range, getunr is used to retrieve the unresolved cross sections
as a function of σ0 , and Eq. 261 is used to compute the weighting flux. In both
cases, the data at each energy are stored as the 1+(lord+1)*nsigz components

254 NJOY2016
8 GROUPR LA-UR-17-20093

e, phi(il,iz),

where il runs from 1 to lord+1 and iz runs from 1 to nsigz.

If the flux calculator option was requested, genflx sets up the parameters
for the calculation and requests needed storage space. Next, the cross section
retrieval routines gety1 and gety2 are set up to return total and elastic cross
sections from the PENDF tape. A lower energy limit, felo, is chosen, and the
cross sections are read into storage until a maximum energy, fehi, or a maximum
number of points, nemax, is reached.
The slowing-down equation, either Eq. 272 or Eq. 276, is then solved from the
break energy down to felo. The scattering source from energies above the break
is based on the NR approximation. When the calculation is finished, fluxes from
10−5 eV to felo are written to the scratch file using loada in the same format
used for the Bondarenko option. From felo to the energy break point, fluxes
are transferred from memory to the loada file. Finally, above the break point,
fluxes are computed and saved using the Bondarenko model.
Subroutine init is used to set up the number of σ0 values, secondary en-
ergy groups, and Legendre components for each combination of mfd and mtd. If
mfd=8, a special copy of File 6 is made for use in getaed. The list of reactions
to be self-shielded can be changed if desired. The number of secondary groups
ng helps determine the storage required for the accumulating group integrals
(see allocatable array ans) in groupr. For simple cross section vectors, ng=2.
The nz*nl flux components are stored first, followed by the nz*nl cross section
components. When all the panels for one group have been processed, dividing
position 2 by position 1 gives the group-averaged cross section. For ratio quan-
tities like ν and µ (mtd=251-253, 452, 455, 456), ng is 3. Once again, the flux
components are stored first, followed by the nl*nz components of ratio*sigma,
followed by the components of the cross section. This arrangement allows for
the calculation of group-averaged values of ratio*sigma, ratio, or cross section
by dividing position 2 by 1, position 2 by 3, or position 3 by 1, respectively. For
matrices, ng is set to one more than the number of secondary groups (ngn or
ngg). The nl*nz flux components are stored first, followed by the nl*nz integrals
for each secondary-energy group in turn.
Subroutine panel performs the generalized group integrals using the logic
described in Section 8.11. For most calls to panel, the lower point of each
“panel” was computed as the upper point of the previous panel. Therefore, panel
is careful to save these previous values. However, if the bottom of the panel is
just above a discontinuity, new values of cross section and flux are retrieved.

NJOY2016 255
LA-UR-17-20093 8 GROUPR

Once the values at the lower boundary of the panel are in place, new values for
the top of the panel are retrieved (see flux, sig). If the top of the panel is
at a discontinuity in σ or at a group boundary, the energy used is just below
the nominal top of the panel. “Just below” and “just above” are determined by
rndoff and delta. For maximum accuracy, these numbers should be chosen
such that rndoff>1, delta<1, and rndoff*delta<1 for the precision of the
machine being used.
For simple average cross sections, the integrals of σ×ϕ and ϕ are computed for
the panel using trapezoids. This is justified by the linearization of σ; the value
of σ×ϕ at the midpoint is too uncertain to justify a more complex treatment.
For two-body scattering, the feed function is far from linear over the panel. In
fact, it can show oscillations as described in Section 8.12. The integral of the
triple product F×σ×ϕ is obtained by Lobatto quadrature of order 6 or 10 using
the quadrature points and weights given in the parameter statements (see qp6,
qw6, qp10, qw10). The cross section and reaction rate are determined at each
quadrature point by interpolation, and the feed function is obtained by getff.
For many reactions, ff will be nonzero for only a certain range of secondary
groups. The value ig1 is the index to the first nonzero result, and ng1 is the
number of nonzero values of ff in the range. Subroutine panel maintains the two
corresponding values iglo and ng to specify the nonzero range of values in array
ans. Finally, the flux and cross section at the top of the panel are transferred to
flst and slst, and control is returned to the panel loop in groupr.
Subroutine displa is used to print cross sections and group-to-group matrices
on the output listing (nsyso). Small values are removed for efficiency. Note
that different formats are used in different circumstances. Infinitely dilute data
are printed without σ0 labels. Isotropic matrices are printed with several final
groups on each line. Delayed neutron spectra are printed using the Legendre
order variable for time groups and with the time constants given on a heading
line.
Subroutine getflx returns nl*nz components of the weighting flux. If nsigz
is 1, the flux is computed using getwtf, and all Legendre orders are taken to be
equal. When self-shielding has been requested, the flux components are obtained
by interpolating between adjacent values retrieved with finda from the scratch
tape written by genflx. The grid energies found on the scratch tape are used
to get enext. The flux is taken to be continuous, so idis is always set to zero.
Subroutine getyld returns the yield needed by getff for fission (MT=452,
455, or 456 from File 1) or radionuclide production yields from File 9 (using

256 NJOY2016
8 GROUPR LA-UR-17-20093

mfd=3zzzaaam). This routine also retrieves the delayed neutron time constants
when mtd=455. Tabulated yields are obtained by interpolation using the utility
routine terpa. Polynomial data are expanded by direct computation.
Subroutine getsig returns nl*nz components of the cross section using point
data from the PENDF tape and self-shielded unresolved data, if present, from
getunr. The routine starts by adjusting MF and MT for the special options
(mtd=258-259, mfd=3zzzaaam, etc.) and locating the desired section on the
PENDF tape. Subroutine getsig is then called for each desired energy value
E in increasing order. For mtd=258 or 259, the appropriate velocity or lethargy
is computed from E and returned. In the more general cases, gety1 is used to
retrieve the pointwise cross section from the PENDF tape, and getunr is called
to replace this value with self-shielded unresolved cross sections if necessary.
The unresolved cross sections are handled using stounr and getunr. Subrou-
tine stounr locates the desired material and temperature in MF=2, MT=152 of
the PENDF tape, and then it copies the data into the global allocatable storage
array unr. The σ0 grid on the PENDF tape does not have to agree with the
list requested for groupr, but a diagnostic message will be printed if they are
different. However, the ntemp values of temperature requested by groupr must
agree with the first ntemp temperatures on the PENDF tape, or a fatal error will
result. Subroutine getunr checks whether E is in the unresolved energy range
and whether MT is one of the resonance reactions. If so, it locates the desired
interpolation range in the unr array, and interpolates for the self-shielded cross
sections. If this is an energy range where resolved and unresolved ranges over-
lap, the resolved part is added to the background σ0 before interpolation. The
subroutine terpu is used for interpolating in the unresolved cross section tables.
The special value MT=261 is used to select the ℓ=1 component of the total cross
section. Note that getsig also uses this ℓ=1 value for ℓ=2, 3, . . . .
Subroutine getff returns the feed function ff using different portions of the
coding for different options. The first section is used for cross sections and ratio
quantities. The same yield yld is returned for every ℓ-component in ff.
The second section of getff is for neutron continuum transfer matrices. The
yield is either determined from MT [for example, yld=2 for MT=16, the (n,2n)
reaction], or it is obtained using getyld for fission. Next, the angular distri-
bution is obtained using getfle (see F in Eq. 328), and the secondary energy
distribution is obtained using getsed (see g in Eq. 328). Finally, the product of
the three factors is loaded in ff. Note that the range of groups returned extends
from iglo=1 to the highest nonzero result, for a total of ng groups. Since ff for

NJOY2016 257
LA-UR-17-20093 8 GROUPR

these reactions is a smooth function of incident energy, nq is set to zero, and no

additional quadrature points will be used in panel.
The third section of getff is for gamma production matrices. The photon
yields are obtained using getyld. In general, there are nyl different yields, each
one corresponding to a different discrete gamma ray, or to the continuum. The
angular distributions for these gamma rays are obtained using getgfl (most
ENDF/B photons are given as isotropic). The energy distribution for the con-
tinuum (if any) is obtained by getsed. Now, the code loops through the pho-
ton group structure placing each discrete photon in the appropriate group and
adding in the continuum part. During this loop, the range of nonzero values
(iglo, ighi) is determined. Finally, the nonzero values are packed into ff, and
iglo and ng are returned to describe the distribution. Again, nq=0 is used.
The fourth section of getff handles two-body scattering, either elastic or
discrete-level inelastic, and both neutrons and charged particles. First, subrou-
tine parts is called to set up the particle type, A′ value, and the scattering law
for reactions that use File 4. Then getdis is called to finish the processing.
The fifth section handles thermal-neutron scattering. The bulk of the work
is done by getaed, which is discussed below. Subroutine getff takes the output
of getaed and packs it into the final result, ff.
The last two sections of getff are used to process energy-angle distributions
from ENDF-6 format sections of File 6. The subroutine simply calls getmf6.
Subroutine parts is used to set up particles for reactions that use File 4.
Different branches are used for ENDF-6 and earlier ENDF versions because the
MT numbers used for charged-particle discrete levels have been changed. For
example, proton levels use MT=600-649 in ENDF-6 libraries, but MT=700-719
in earlier libraries. If mfd=3, no action is taken. For other mfd values, the routine
determines zap, aprime, and law. The result depends on whether File 4 or File
6 data are to be used.
Subroutine getmf6 is used to compute the feed function for reactions repre-
sented in File 6. As is common with NJOY subroutines, it is called with ed=0 to
initialize the subroutine. The first step is to locate the desired section of File 6
on the input ENDF tape. It then sets the parameter zad (for ZA desired) based
on the input value mfd and searches through the section for the desired subsec-
tion. If this subsection has law=4, it defines a recoil particle; the code backs
up to the first subsection in the section, which is assumed to be the subsection
describing the emitted particle. The subroutine has now arrived at statement
number 140, where it decides whether to branch to special coding for two-body

258 NJOY2016
8 GROUPR LA-UR-17-20093

reactions (statement number 500) or phase-space reactions (statement number

150). All the other options have a TAB2 record at this point that defines the
incident energy grid. For law=1, the data for the first incident energy are read
in and converted to the Lab frame by cm2lab. Similarly, for law=7, the data are
read in and converted to a law=1 format by ll2lab.
For a normal entry, getmf6 interpolates for the particle yield for the current
particle. For all of the laws except 2-5 (discrete two-body laws), it then checks to
see whether the energy ed is in the current panel (or if this is the first time in the
first panel). If not, it moves the high data to the low position and reads in new
high data. Of course, it also converts the new high data to the correct form with
cm2lab or ll2lab. Once ed is in the current panel, the code reaches statement
number 300, which sets up the loop over secondary energy by initializing f6lab.
This loop uses a grid that consists of the epnext values returned by f6lab
and the group bounds eg(i). The actual integral over EL′ inside each group
uses trapezoidal integration. The last step for laws 1, 6, and 7 is to add in
the contributions to the feed function from discrete energies in File 6. For the
discrete two-body laws (2-5), the routine goes through statement number 500,
which simply calls getdis. For all laws, the last step is to check the normalization
of the feed function. If the error is large, a message is printed out. In any case,
the results are adjusted to preserve exact normalization.
Subroutine cm2lab is used to convert a CM distribution starting at inow in
cnow into a Lab distribution, which will be stored starting at jnow. This routine
generates a grid for the Lab secondary energy EL′ by adaptive reconstruction.
The reconstruction stack is first primed with pLℓ (E, 0) as computed by f6cm
and pLℓ (E, Enext
′ ) (also from f6cm), where the value Enext
′ is the value epnext
returned by the first call to f6cm. This panel is then divided in half, and the value
returned by f6cm is compared with the linear interpolate. If they agree within
0.5% (see tol), this panel is converged. Otherwise, it is subdivided further.
When convergence has been achieved for this panel, a new panel is chosen, and
it is subdivided to convergence. When the entire range for EL′ has been processed,
the final parameters are loaded into the new File 6 record for pLℓ (E, EL′ ) (which
starts at jnow in cnow), and the routine returns to getmf6.
Subroutine f6cm is used to compute the Legendre coefficients of the double-
differential scattering function pℓ (E→E ′ ) in the Lab system from data given in
File 6 in the CM frame. The memory area cnow contains the raw data. It is
necessary to call the routine ep=0 for each new value of E. Thereafter, values
of ep can be requested in any order. This is required by the adaptive scheme

NJOY2016 259
LA-UR-17-20093 8 GROUPR

used to generate the Lab E ′ grid in cm2lab. The conversion uses Eqs. 361 - 366.
On a normal entry with EL′ >0, the code computes the lower limit of the integral
µmin . It then sets up an adaptive integration over a panel starting at µ=1. The
value for EC′ is computed using

EC′ = (1 + c2 − 2cµL )EL′ , (375)

which is based on Eq. 362, and either f6ddx or f6psp is called to compute
pC (E, EC′ , µC ) and epnext. The lower limit of the integration panel is then
computed by setting EC′ equal to epnext and solving for the CM cosine using

µL − c
µC = p . (376)
1 + c2 − 2cµL
Of course, the larger of this result and µmin is used as the lower bound of the
panel. Therefore, the adaptive integration operates on a nice continuous func-
tion. It continues subdividing the µC grid in this panel until convergence is
achieved within 0.5% for all Legendre orders. The lower bound of this panel
becomes the upper bound of a new panel, and a new lower bound is selected as
above. The integration is carried out over successive panels until µmin is reached.
When the calculation of the continuum part of pLℓ is finished for these values
of E and EL′ , the routine checks for possible contributions from delta functions.
Next, the routine scans through the computed coefficients and zeros out any
small ones that may just be noise. Finally, it chooses a value for epnext and
returns.
Function f6ddx is used to compute the double-differential scattering function
f (E→E ′ , ω), where secondary energy E ′ and scattering cosine ω are in the CM
system. On entry, cnow contains the File 6 data for a particular value of E, and
f6ddx must be called once with ep=0 for each new value of E. Thereafter, ep
values can be requested in any order (this is required by the adaptive scheme
used to convert to the Lab system in f6cm). For a normal entry with ep>0, the
code searches for the panel in the data that contain the requested value of E ′ .
If lang=1 for this subsection of File 6, the data are already given as Legendre
coefficients, and the code simply interpolates for the desired results. If lang=2,
the data use the Kalbach-Mann scheme for representing the energy-angle distri-
bution. This routine includes both the original Kalbach-Mann representation[59]
and the newer Kalbach representation[60]. It has been set to use the latter by
the parameter statement k86=1. The code interpolates for the model param-
eters at E ′ and computes the desired answer with the model’s formulas. The

260 NJOY2016
8 GROUPR LA-UR-17-20093

ENDF-6 format also allows CM distributions given as values tabulated versus

scattering cosine (see lang=11-15). Note that there is a “stub” to take special
action at low energies. It is currently disabled by the statement efirst=0.0, but
it may be used sometime in the future to account for the fact that the low-energy
√
dependence of the scattering function must vary as f (E ′ ) = c ∗ E ′ .
Function bach computes the Kalbach-86 a parameter, which depends on the
neutron separation energy for the target izat using the liquid-drop model with-
out pairing and shell terms. The formula used describes the reaction a + B →
C + d:

Esep = 15.68 AC − AB
(NC − ZC )2 (NB − ZB )2

− 28.07 −
AC AB
2/3 2/3

− 18.56 AC − AB
(NC − ZC )2 (NB − ZB )2

+ 33.22 4/3
− 4/3
AC AB
ZC2 ZB2

− 0.717 1/3 − 1/3
AC AB
Z 2 Z2

+ 1.211 C − B , (377)
AC AB

where A stands for atomic weight, Z for charge number, and N for neutron
number. Note that, even for reactions like (n,2n), C is the residual nucleus
resulting from the removal of one particle, d; it is not necessarily the real physical
residual nucleus for the reaction. If the target izat is an element, Esep has to
be computed for some dominant isotope in the element. Dominant isotopes
are assigned in this routine for some materials that often appear as elements
in ENDF evaluations; if the particular target required does not appear, a fatal
error message is issued. The user will have to add a line to the routine for the
material and reassemble the code.
Subroutine ll2lab converts File 6 from Law-7 format to Law-1 format using
the laboratory Legendre representation. Law 7 represents the double-differential
scattering distribution f (E→E ′ , µ) by giving a series of tables of f (E→E ′ ) for
series of µ values. On entry, the entire Law-7 section is stored in c starting at
inow. The code loops through a set of E ′ values chosen to be the union of all the
E ′ grids for all the different µ values. For each point on this union grid, the code
interpolates for all the corresponding f (µ) values, and it uses them to compute

NJOY2016 261
LA-UR-17-20093 8 GROUPR

nl Legendre coefficients fℓ (E ′ ). After completing the calculation for a new value

of E ′ , it checks the normalization of the result, and then it checks back to see
whether the previous value is still needed to represent the curves fℓ (E ′ ) within a
tolerance of 0.5%. The results are written in c starting at jnow. They have the
Law-1 Legendre form, namely, sets of values E ′ , f0 (E ′ ), f1 (E ′ ), . . . , fNL−1 (E ′ ),
given for a series of E ′ values starting with zero.
Subroutine f6lab is used to return the Legendre coefficients of the double-
differential scattering function fℓ (E→E ′ ) for particular values of E and E ′ in the
Lab system (see e and ep) from a part of File 6 in Law-1 format. Since Law-7
sections have been converted to Law-1 format using ll2lab, and since CM sec-
tions in Law-1 format have been converted to the Lab frame using cm2lab, the
only two roles left for this subroutine are interpolation to the desired values of
E and E ′ and preparation of Legendre coefficients for sections that use the File-
6 variant with law=1 and lang=11-15 (laboratory distributions tabulated) vs.
µ. Only the continuum portion of the distribution is processed here; any delta
functions given in File 6 must be handled separately. The routine is initialized
by calling it with ep=0. Data for the two incident energy values that bracket E
are already present in clo and chi. Then the routine extracts various param-
eters from the c array and prepares the variables used to control interpolation.
These variables are complicated, because this routine handles three different in-
terpolation schemes: “cartesian”, “unit base”, and “corresponding point” (see the
ENDF-6 manual[9] for more details). For a normal entry, the code searches the
data in clo and chi for the intervals containing E ′ . If lang=1, it performs a
two-dimensional interpolation for the nl coefficients at E and E ′ . For lang>1,
it computes the Legendre coefficients from the File 6 data, and then does the
two-dimensional interpolation. Finally, it computes epnext and returns.
Subroutine getdis is used to compute the feed function for elastic or discrete
inelastic scattering of neutrons or charged particles using either File 4 or File 6
data. First, the angular distribution is retrieved with getfle, and an appropriate
quadrature order is selected using Eq. 319. Then, a group loop from low energy
to high energy is used to compute the ω1 and ω2 limits of Eq. 317, and the range
(ω1 , ω2 ) is subdivided using the appropriate Gauss-Legendre quadrature points
(see qp4, qp8, qp12, and qp20). The function f (E, ω) is computed at each of
these quadrature points using Eq. 318 and the angular distribution previously
returned by getfle. The laboratory cosine at the quadrature point, µ ω , is
 

computed using Eq. 314. Finally, the integrand of Eq. 317 is multiplied by the
appropriate quadrature weight (qw4, qw8, qw12, or qw20) and added into the

262 NJOY2016
8 GROUPR LA-UR-17-20093

accumulating integral. This continues until ω2 = 1. The nonzero values of ff

and the parameters iglo and ng are then complete for this value of E.
The next step is to determine enext based on the next critical point as given
by Eqs. 321 - 324. Special cases are used for elastic scattering to avoid numerical
problems. Note that the discontinuity flag idisc is set for critical points. The
nq variable is also set to force panel to subdivide the initial-energy integration.
Subroutine getfle retrieves or computes the Legendre coefficients for the
angular distribution of a reaction at incident energy E. When called with e=0,
getfle requests storage for the raw data, reads in the File 4 information for
the first two incident energies on the file, and then uses getco to retrieve the
corresponding coefficients. On subsequent entries with e>0, getfle simply inter-
polates for the desired coefficients. When e exceeds the upper energy in storage,
the values for the upper energy are moved to the lower positions, and new up-
per values are read and converted to coefficients. An isotropic distribution is
returned if e is outside the range of the angular data from File 4.
Subroutine getaed retrieves angle-energy data for thermal scattering reac-
tions. For coherent elastic scattering, the routine reads through the cross section
on the PENDF tape using gety1 and locates the Bragg edges. On each subse-
quent call to getaed, the Legendre components of the cross section are computed
using Eq. 309. For incoherent elastic scattering, the routine is first initialized by
reading in the raw data for the first energy. On subsequent entries, a test is made
to see whether e is in the range elo to ehi. If not, the high data are moved
to the low positions, and new high data are read. The Legendre components
are then computed using Eq. 311. For incoherent inelastic scattering, getaed
is initialized by reading in the raw data for the first two incident energies. On
subsequent entries, the subroutine checks to see whether e is between elo and
ehi. If not, the data for ehi are moved to the low positions, and new raw data
are read from File 6 and binned. Once the correct data are in place, the desired
energy-angle distribution is computed by using a combination of interpolation
along lines of constant energy transfer and unit-base interpolation.
Subroutine getgfl returns the Legendre coefficients for the angular distri-
butions for all discrete and continuum photons for a reaction simultaneously.
When called with ed=0, the routine reads File 14 into scratch storage and finds
the starting location for the subsection describing each photon. On subsequent
entries with ed>0, getgfl sets up a loop over the ng photons on this section
of File 14. For each photon, it searches for the energy panel that contains ed,
uses getco to retrieve or compute the Legendre coefficients at the upper and

NJOY2016 263
LA-UR-17-20093 8 GROUPR

lower File 14 energies, and interpolates for the desired coefficients at ed using
terp1. Since most ENDF photons are represented as isotropic, a special short-
cut calculation is provided for that case. Isotropic results are also returned if ed
is outside the range of the data in File 14.
Subroutine getco is used by both getfle and getgfl to retrieve or compute
Legendre coefficients from data in File 4 or File 14 format. The user can request
output in either the Lab or CM system, and the raw data can be either Legendre
coefficients or tabulated probability versus emission cosine. However, if the raw
data are in the laboratory system, CM coefficients cannot be produced. If the
raw data are already in the form of coefficients in the desired system, getco
simply checks for the maximum Legendre order needed using a tolerance of
toler=1e-6 and returns the coefficients in fl and the order in nl. If coordinate
conversion is required, or if the raw data are tabulated, getco sets up the integral
over cosine using Gauss-Legendre quadrature of order 20 (see qp and qw). The
scattering probability for the quadrature point is computed from the coefficients
or obtained by interpolating in the tabulation. The Legendre polynomials in
the desired reference system are then computed. If the raw data are in the CM
system (ω) and the result is to be in the Lab system (µ), the desired polynomials
are Pℓ (µ ω ); otherwise, the quadrature angle is used directly to compute the
 

polynomials. Once the coefficients have been computed, they are checked using
toler to determine the maximum order, nl, and the results are returned in fl.
Subroutine getgyl is used to retrieve the yields for all photons emitted in a
specified reaction simultaneously. The raw data are obtained from the ENDF/B
tape as either yields (MF=12) or production cross section (MF=13). In the lat-
ter case, getgyl actually returns the fraction of the total yield assigned to each
photon. The cross section returned by getsig is the total photon production
cross section from MF=13 on the PENDF tape, which makes the resulting in-
tegral correct. Using the normal GROUPR procedure, getgyl is initialized by
calling it with ed=0. The entire File 12 or File 13 is read into scratch storage,
and the starting location for each subsection is determined. On subsequent en-
tries (ed>0), the routine loops over the nyl photons found, and uses terpa to
compute the yield at ed. If this is a primary photon, a discontinuity is set up at
the energy where the photon will change groups. For MF=12, the calculation is
finished. For MF=13, the numbers calculated above are converted to fractions of
the total yield by dividing by the total production subsection from File 13. This
routine does not handle ENDF/B transition probability arrays directly, because
they will have been converted to File 12 yields by conver.

264 NJOY2016
8 GROUPR LA-UR-17-20093

Subroutine conver converts ENDF/B evaluations to a standard form. If

transition probability arrays were used in File 12, they are converted to yields
and written back into File 12. If a section with MF=1 and MT=456 is missing
from the evaluation, a copy of MT=452 is added to the tape as MT=456. In
addition, a second copy of the modified tape is made on unit nscr. While conver
is reading through the tape, lists of the reactions in File 4, File 12, and File 13
are written to to global arrays for use by the automatic reaction selection logic in
nextr. For ENDF-6 evaluations containing File 6, the routine scans through File
6 looking for sections that produce neutrons, photons, or charged particles. The
MT numbers for these sections are stored into mf6, mf18, and mf6p. The routine
also checks for sections of File 4 containing charged-particle angular distributions
and records their MT numbers in mf4r. Finally, if the section MF=6/MT=2
contains charged-particle elastic scattering information given using the nuclear-
plus-interference format, it is converted into the residual-cross-section format for
getdis.
Subroutine getsed returns the secondary-energy distribution for neutrons or
continuum photons for all groups simultaneously. Both tabulated and analytic
functions are handled. getsed is initialized for a particular reaction by calling it
with ed=0. First, scratch storage is allocated, and all the subsections are read
in. Analytic subsections are left in their raw form, but tabulated subsections
are averaged over outgoing energy groups for each of the given incident energies.
The array loc contains pointers for each subsection. On subsequent entries
(ed>0), getsed loops over the subsections for this reaction. It first retrieves
the fractional probability for the subsection using terpa. If an analytic law
is specified, anased is used to compute the group integral for each secondary-
energy group. Each integral is multiplied by the fractional probability for the law
and accumulated into sed. For tabulated data, the routine simply interpolates
between the two values for the group integrals using terp1, and accumulates
them into sed. Note that various restrictions on the ordering of subsections
and prohibition of multiple tabulated subsections needed for earlier versions of
GROUPR are no longer required. Upscatter is not allowed in secondary-energy
distributions except for fission or photon production. If found, it is put into the
“in-group” (g ′ =g).
Subroutine anased is used to calculate the integral from e1 to e2 for one of
the analytic laws (see Eqs. 331 - 360). The routine uses the SLATEC version
of the reduced complementary error function from the NJOY2016 math module.
The resulting integral is returned in g.

NJOY2016 265
LA-UR-17-20093 8 GROUPR

Subroutine hnab is used to compute the special functions required for analytic
law 11, the energy-dependent Watt spectrum. The method used is described in
the BROADR chapter of this manual.

8.20 Error Messages

The fatal error messages and warning messages from GROUPR are listed below,
along with recommended actions to recover from the problem.

error in groupr***unable to find mat=–-, t=–-.

Check for input error or wrong PENDF tape mounted.
error in groupr***photons not allowed with igg=0.
In order to produce photon data, a photon group structure must be re-
quested. Check the input on card 2.
error in groupr***illegal mfd.
Check input mfd; legal values are 3, 5, 6, 8, 12, 13, 16, 17, 18, 21-26, 31-36,
and the isotope production values like Xzzzaaam for X = 1, 2, 3 or 4.
message from groupr–-auto finds no reactions for mf=–.
An automatic reaction selection card of the form “mfd” was given in the
input, but the ENDF and PENDF tapes do not contain any sections that
would produce the desired cross sections or matrices.
error in groupr***unable to find next temp.
The current material ended before the requested temperature was found.
error in ruinb***illegal ismooth.
Check the input, the value for the smoothing option must be either 0 or 1.
error in gengpn***read-in group structure is out of order.
Group structures must be given in ascending energy order.
error in gengpn***illegal group structure.
Check input; current legal values are 1 through 34.
error in gengpg***illegal group structure.
Check input; current legal values are 1 through 10.
error in genwtf***exceeded storage reading user weight function
See the allocatable array tmp with length ntmp=10000.
error in genwtf***illegal weight function.
Check input; current legal values are −12 through +12.
error in genflx***total not defined over energy range.
A complete total cross section is needed for self-shielding. This means that
“dosimetry” and “activation” tapes, which normally give only a few key
reactions, can only be processed using nsigz=1 (infinite dilution).

266 NJOY2016
8 GROUPR LA-UR-17-20093

error in getfwt***temperature ...

The requested temperature could not be found while searching the tape.
NJOY uses a tolerance of 1e-6 for this purpose.
error in getfwt***e outside range of data.
A premature end-of-file was found on the input flux tape when using the
iwt=0 option. Check to be sure the tape was the output from a legal flux
calculator run.
error in getfwt***requested e is out of order.
The cause could be an improper input tape. Check as above.
error in panel***elo.gt.ehi.
This indicates some error in the energy grid for panel. It usually oc-
curs if rndoff and delta are incorrect for your machine. Make sure that
rndoff>1, that delta<1, and that the product rndoff*delta< 1 when
evaluated on your machine (for example, 1.00000001 is not greater than
unity on a 32-bit machine).
error in panel***bad nq in panel
The nq parameter can be 2, 6, or 10 with the currently installed quadra-
tures.
message from panel–-thermal range problem at ...
NJOY expected scattering to a specific group but found another instead.
error in getyld***illegal lnd.
The maximum number of time groups is 8. See global array dntc(8).
error in getyld***unable to find nuclide for iza=... lfs=...
Unable to find requested nuclide production yield.
error in getsig***illegal mt.
Check input for mtd.
error in getsig***can’t find mf,mt,lfs ...
Check input for mfd greater than 10000000.
message from stounr–-no unresolved sigma zero data....
This message probably means that UNRESR or PURR was never run for
this material. Infinitely dilute values will be used.
message from stounr–-sigma zero grids do not match....
The unresolved calculations will probably work best if the σ0 grid in GROUPR
matches the one in UNRESR or PURR. However, this is not necessary.
getunr will interpolate to get values on the GROUPR grid from the UN-
RESR or PURR grid. A message is issued in case this isn’t what the user
really intended.
error in stounr***storage exceeded.
There is not enough storage for unresolved cross section data on PENDF
tape. The allocatable array tmp has length ntmp=10000.

NJOY2016 267
LA-UR-17-20093 8 GROUPR

error in stounr***cannot find temp=–-

The list of temperatures requested for the GROUPR run must agree with
the first ntemp temperatures on the PENDF tape. Check your BROADR
and UNRESR or PURR runs.
message from getunr–-Warning, negative URR cross sections found
NJOY found negative cross sections for the unresolved energy range from
UNRESR.
error in getff***do not know how to handle mf,mt ...
The getff routine branches to different blocks of coding for different com-
binations of mfd and mtd, but if no appropriate branch is found for the
current values, this error message is issued. It probably indicates an error
in the evaluation.
error in getmf6***desired particle not found.
The outgoing particle for a group-to-group matrix is implied by the value
of mfd (for example, protons for mfd=21). This message means that the
section of File 6 requested with mfd and mtd does not contain a subsection
that produces that particle. Check the user input. This message should
not occur with automatic reaction selection.
error in getmf6***illegal law.
The value of the law parameter is greater than 7. This implies an error in
the evaluation.
error in getmf6***too many subsection energy points.
Limited by the parameter maxss=500.
error in getmf6***storage exceeded.
See the allocatable array temp with length ntmp=990000.
message from getmf6–-bad grids for corresponding point ...
Corresponding-point interpolation won’t work correctly unless the two grids
above and below the point of interest have the same number of points. This
message means that there is an error in the evaluation.
error in getmf6***too many subsections for one particle.
We currently allow for no more than three. See iyss(3), izss(3), and
jss(3).
message from getmf6–-there are multiple subsections in mf6
This warning message is issued when a specific particle has multiple sub-
sections in an mf6 section.
error in cm2lab***storage exceeded.
This means that the allocatable array tmp with length ntmp=990000 in
subroutine getmf6 has run out of space.
message from cm2lab–-lab normalization problem mt=... e=...
This warning message is issued when the integral after normalisation to the
lab system is different from 1 by more than 1%.

268 NJOY2016
8 GROUPR LA-UR-17-20093

error in f6ddx***illegal lang.

The value of lang for tabulated angular distributions must be in the range
11-15.
message from cm2lab–-vertical segment(s) in distribution
There appears to be a jump in the energy dependent multiplicity for the
outgoing particle at a given incident energy. This message is only given
when Kalbach-Mann systematics are used. This is most likely an evaluation
problem.
error in bach***dominant isotope not known.
The calculation of the neutron separation energy needed for the Kalbach
model for particle energy-angle distributions needs a value for the dominant
isotope in an element. It will have to be added to the code. The same
problem will occur with the parallel routines in HEATR and ACER.
error in ll2lab***storage exceeded.
This means that the allocatable array tmp with length ntmp=990000 in
subroutine getmf6 has run out of space.
error in f6cm***nl>mxlg
The current limit is mxlg=65.
error in f6ddx***nl>mxlg
The current limit is mxlg=65.
error in f6lab***illegal lang.
The value of lang must be equal to 1 or be in the range 11–15 for tabulated
angular distributions.
error in f6dis***illegal lang.
The value of lang must be equal to 1 or 2, or be in the range 11–15 for
tabulated angular distributions.
error in getdis***illegal nqp
The allowed quadrature orders are 4, 8, 12, and 20.
error in getfle***desired energy above highest energy given
Should not occur for well-constructed ENDF files. Check the evaluation to
be sure File 3 and File 4 are consistent.
message from getfle–-lab distribution changed to cm for mt=...
Angular distributions for two-body reactions are supposed to be given in the
CM frame by ENDF conventions. Some old evaluations for heavy materials
violate this rule; changing to the CM frame has little effect on the answers.
error in getaed***thermal mf6/law7 not coded
This subroutine can’t handle the E-µ-E ′ ordering option provided by THERMR.
error in getaed***storage exceeded.
This refers to the allocatable array aes with length maxaes=200000.

NJOY2016 269
LA-UR-17-20093 8 GROUPR

error in getgfl***too many gammas.

See the parameter maxgfl=500.
error in getgfl***storage exceeded.
See the parameter ntmp=10000.
error in getgfl***desired energy at highest given energy.
This problem should not occur in a well-constructed ENDF file. Check
Files 3, 12, 13, and 14 for consistency.
error in getco***limited to 64 legendre coefficients.
See nlmax=65 and P(65).
error in getco***lab to cm conversion not coded.
The need for this type of conversion rarely occurs on the current ENDF
evaluations, because CM is consistently used for two-body reactions, and
the laboratory frame is consistently used for continuum reactions. There are
a few exceptions for the heavy isotopes, where CM and lab are essentially
equivalent, but they were errors when the files were generated.
error in getgyl***lo=2 not coded.
This message should not occur, because any transition probability arrays
on the ENDF tape should have been converted to yields by conver.
error in getgyl***too many gammas.
The current limit is 550 photons. See the parameter nylmax=550.
error in getgyl***storage exceeded.
This refers to the allocatable array tmp with length ntmp=15000.
message from conver–-cannot do complete particle production...
With the advent of the ENDF-6 format, it is possible to make evaluations
that fully describe all the products of a nuclear reaction. Some carry-over
evaluations from earlier ENDF/B versions also have this capability, but
many do not. This message is intended to goad evaluators to improve
things!
message from conver–-gamma production patch made for ...
This patch is used to correct the old ENDF/B-III evaluations for MAT=1149
and MAT=1150 (chlorine and potassium).
message from conver–-mf12, mt ... may be missing
This message indicates that the discrete photon data in mf12 for this mt
number may be missing or be incomplete.
error in conver***nnth too large
See mxnnth=350.
message from conver–-skipping new mf6/mt18 multiplicity section
Starting with ENDF/B-VIII.0, fission neutrons and photons can now be
described using probability functions for emitting 0, 1, 2, ... particles per
fission. This data can currently not be included in any multigroup or con-
tinuous energy format so the data is simply skipped.

270 NJOY2016
8 GROUPR LA-UR-17-20093

error in conver***too many lo=2 gammas.

The LO=2 processing uses a set of allocatable arrays that are sized using
lmax=500. That number can be increased freely, if necessary.
error in conver***storage for fission nu exceeded.
The storage in the allocatable array nu is sized using nnu=6000, which can
be increased freely.
error in conver***xxxx to big.
The automatic processing of reactions is controlled by lists stored in global
arrays like mf4 or mf6p. This error occurs when one of the particular in-
dexes, indicated by “xxx” exceeds maxr1=500. The arrays for MF=10 (see
imf10) have a limit of maxr2=500.
error in getsed***too many subsections.
The current limit is 20. See the parameter nkmax=20.
error in getsed***storage tmp exceeded.
The input ENDF data are stored in the allocatable array tmp with length
ntmp=50000. An integer will be included in this message to indicate where
exactly in the source code this message was issued.
message from getsed–-corresponding point interpolation ...
The interpolation schemes corresponding to the range 11–15 are not sup-
ported.
message from getsed–-upscatter correction....
This reaction should not have upscatter. The error is placed into the in-
group element.
error in anased***illegal lf.
Legal values are 5, 7, 9, 11, and 12.
error in f6psp***3, 4, or 5 particles only.
Phase-space formulas for 3, 4, or 5 particles are provided in this routine.
Check for an error in the evaluation.

NJOY2016 271
LA-UR-17-20093 8 GROUPR

272 NJOY2016
9 GAMINR LA-UR-17-20093

9 GAMINR
The GAMINR module of NJOY is designed to produce complete and accurate
multigroup photoatomic (but not photonuclear) cross sections from ENDF/B-IV
and later data[68], including the newer formats developed for ENDF/B-VII. To-
tal, coherent, incoherent, pair-production, and photoelectric cross sections can
be averaged using a variety of group structures and weighting functions. The
Legendre components of the group-to-group coherent and incoherent scattering
cross sections are calculated using the form factors[69] now available in ENDF/B.
These form factors account for the binding of the electron in its atom. Conse-
quently, the cross sections are accurate for energies as low as 1 keV. GAMINR
also computes partial heating cross sections or kinetic energy release in materials
(KERMA) factors for each reaction and sums them to obtain the total heating
factor. The resulting multigroup constants are written on an intermediate inter-
face file for later conversion to any desired format. Photonuclear reactions such
as (γ,n) are not computed by this module.
GAMINR differs from the previously used GAMLEG code[23] in the following
ways:

• Coherent form factors are processed thereby allowing higher Legendre com-
ponents of the coherent scattering cross section to be produced. GAMLEG
processed the P0 cross section only.
• Incoherent structure factors are processed giving accurate results at low
energies where the Klein-Nishina formula fails.
• Heat-production cross sections (KERMA factors) are generated.
• Variable dimensioning and dynamic storage allocation allow arbitrarily com-
plex problems to be run.
• GAMINR is much slower than GAMLEG since charge scaling of the inco-
herent matrix can no longer be used at all energies.
This chapter describes the GAMINR module in NJOY2016.0.

9.1 Description of ENDF/B Photon Interaction Files

In the ENDF/B-IV and later photon interaction files, the coherent scattering of
photons by electrons is represented by

3σT
σC (E, E ′ , µ) dE ′ dµ = (1 + µ2 ) |F (q, Z)|2 δ(E − E ′ ) dE ′ dµ , (378)
8

where E is the energy of the incident photon, E ′ is the energy of the scattered

NJOY2016 273
LA-UR-17-20093 9 GAMINR

photon, µ is the scattering cosine, σT is the classical Thomson cross section

(0.66524486 b), Z is the atomic number of the scattering atom, and F is the
atomic form factor. The coherent form factor is a function of momentum transfer
q given by
r
1−µ
q = 2k , (379)
2
where k is the energy in free-electron units (k = E/511003.4 eV), but F is
actually tabulated versus the quantity x = 20.60744q. The coherent form factor
is given as MT=502 in File 27.
Incoherent scattering is represented by the expression

σI (E, E ′ ,