© J. Collin et al., Published by EDP Sciences, 2025

1. Introduction

Nuclear inventory analysis is essential for various nuclear applications, including reactor fuel depletion, material activation in nuclear devices, radiation dose assessment, and decay heat calculations. For precise radiological inventory characterization, several computational tools are available, such as FISPACT-II [1], DCHAIN-PHITS [2], or TMX_Bateman – an in-house code developed by the private company Transmutex [3]. The latter is developing an Accelerator Driven System (ADS). For their research and development activities involving spent-fuel reprocessing and decay calculations, a dedicated tool as well as comprehensive and well validated nuclear data are essential. This work presents the latter aspect.

Attempts have been made to create test libraries at the destination of ADS benchmarking studies, such as ADS-2.0 [4] and ADS-HE [5]. Current efforts are more focused on the neutron cross-sections. The ADS-2.0 proposes evaluated neutron cross-sections for 156 target nuclides, whereas the ADS-HE interest was to expand the energy range beyond the usual 20 MeV up to 1 GeV of such cross-sections with, for now, ²⁰²Hg, ²⁰⁸Pb, ²⁰⁹Bi, ²³²Th, ²³⁵U, ²³⁸U, ²³⁷Np, ²³⁹Pu, ²⁴²Am and ²⁴⁵Cm. From these incident neutron data, 743 isotopes can be produced, accounting for neither the decay by-products nor the fission fragments. With a proton beam up to 1 GeV [6], one can expect the production of a high number of excited states, that are not sufficiently covered in current decay databases, see overview in the latter section.

Indeed, the aforementioned Bateman solvers rely on individual nuclide data files to specify their respective nuclear properties. However, current databases exhibit limitations in fully accounting for all isomeric states, particularly in the context of fission processes, or in the distribution of all relevant properties (e.g. energy distributions of secondary particles).

Several national and international collaborations are actively compiling and evaluating decay characteristics–including decay channels, branching ratios, and available energies. Notable examples include the Decay Data Evaluation Project (DDEP) [7], which provides high-quality nuclear data for a curated selection of nuclides (228 as of February 2025), and the Evaluated Nuclear Structure Data File (ENSDF) [8], offering recommended nuclear structure and decay data for all known nuclides through systematic evaluations of entire mass chains. Additionally, the NUclear DataBASE (NUBASE) and its associated Atomic Mass Evaluations (AME) [9–11] catalogue properties of experimentally confirmed nuclides alongside extrapolated predictions, forming the most extensive nuclear database currently available.

Although, decay data are released in various formats, depending on their use, the two most widely adopted are ENSDF (favoured by nuclear physicists) and ENDF-6 [12] (designed for efficient computational processing). Notably, the deterministic codes referenced previously primarily use the latter format.

The main sources of decay data are the Joint Evaluated Fission and Fusion library (JEFF) [13], the Evaluated Nuclear Data File B (ENDF/B) [14], the Japanese Evaluated Nuclear Data Library (JENDL) [15] or the Chinese Evaluated Nuclear Data Library (CENDL) [16]. The European Activation File (EAF) [17], the Fusion Evaluated Nuclear Data Library (FENDL) [18], the International Reactor Dosimetry and Fusion File (IRDFF) [19], the United Kingdom Heavy Decay Data/Product Decay Data (UKHEDD/UKPADD) [20], the Waste Incineration Nuclear Data (WIND) [21], amongst others, can also be encountered.

These evaluated databases are updated less frequently than experimental datasets, as they are usually released alongside neutron cross-section evaluations, which is the primary focus of the format originally.

For instance, the ENSDF project, the main provider of decay properties, conducts full evaluations for approximately 220 nuclides annually, with partial updates issued monthly [22]. On average, a given mass chain undergoes reevaluation every seven years [23].

The major libraries can exhibit even longer revision cycles, with seven years between JEFF-3.3 and JEFF-4.0, or between ENDF/B-VIII.0 and ENDF/B-VIII.1, and eleven years between JENDL-4 and JENDL-5. Moreover, these distributions aim to serve a different purpose than broader databases such as NUBASE or the Reference Input Parameter Library (RIPL) [24], which prioritise comprehensiveness.

Furthermore, current databases do not implement all the properties that their format supports, requiring users to address limitations themselves. One of the reason is that user applications vary widely, hence not all limitations are known in advance. And the format evolution introduces new features that require time to be implemented.

For inventory analysis, obtaining precise beta distributions (unlike other secondary particles) requires additional utilities like LogFT [25] or BetaShape [26], as only maximum energy levels are typically provided in the standard ENDF-6 file. Systematic data for internal bremsstrahlung are often unavailable. Yet the contribution of high-energy β-particle bremsstrahlung to the total dose rate can be significant [27]. Consequently, by including it systematically, the Bateman solver needs only to compute the external bremsstrahlung component.

To address these release delays and reduce reliance on external utilities during inventory analysis (especially for TMX_Bateman), we have developed a Python-based toolset that generates an extensive and up-to-date decay database, compiling several nuclear databases. The Global Evaluated Nuclear File Decay sub-library (GENF/D) combines the NUBASE2020 evaluation with the latest ENSDF distribution, accessed via IAEA Livechart API, augmented by calculated quantities from BetaShape for beta spectra, from the GEneral description of Fission observables (GEF) code [28] for fission properties, from the Band-Raman Internal Conversion Coefficients (BrIcc) code [29] for conversion electron data, and from the RadiationReport of the NSDD network [30] for internal bremsstrahlung estimates.

This compiled dataset has been checked against the other evaluated databases, as well as established international benchmarks using the deterministic Bateman equation solver in FISPACT-II.

2. Material and methods

This section provide an overview of the evaluated databases, and in particular NUBASE which is the main source of nuclide characteristics for generating GENF/D library. The additional sources of properties, and code used are presented.

2.1. Overview of decay databases

The evaluated nuclear data, mentioned in the introduction, cover 4219 isotopes (including 3337 ground states) and document 27 distinct decay modes (i.e. described by a unique RTYP, the numerical value used in ENDF-6 format to specify a decay channel). Focusing specifically on the most recent major evaluations – JENDL-5, ENDF/B-VIII.1, and JEFF-4.0 – these provide complete decay data for 3281 ground states and 833 excited isomeric states, characterised by 25 decay modes. The JENDL decay sub-library provides information about 4071 radioisotopes, including 797 excited states, making it the current most extended evaluated ENDF-6 sub-library, see Table 1.

The AME Project, which provides comprehensive nuclear mass data, began consolidating its evaluations into a unified publication starting in 1983. With the 1993 update, the NUBASE database was created to ensure consistent treatment of isomeric states and was first published in 1997 [31]. This database includes not only nuclear masses but also decay modes, branching ratios, and total available energy: all being key properties for radionuclide tracking. Since its initial 1997 release (3891 isotopes, 3010 ground states), NUBASE has expanded significantly, with the 2020 edition containing 5565 isotopes, 3558 ground states, and 43 unique decay modes.

Aforementioned Bateman solvers cannot process nuclides without associated data file. To maximise nuclide coverage, these codes typically employ compiled decay sub-libraries that combine data from multiple major evaluations. For instance, the decay2020 library referenced in Table 1 integrates decay data from JEFF, ENDF/B and JENDL. A comparable approach has been implemented by DCHAIN developers, albeit with different priority schemes for data selection.

Building upon these methodologies, we propose an enhanced strategy: generating a self-consistent database derived from NUBASE. This source is uniquely positioned for this purpose as it offers the most complete compilation of nuclides and isomers within a unified dataset.

2.2. Supporting sources

If one were to convert the NUBASE dataset into ENDF-6 format, the minimal base required to follow an inventory could be achieved. Yet in order to compute the average energy of secondary particle, or decay heat, or to access their energy distribution, an ENSDF-type datasets are necessary. In addition, to estimate the available energy for spontaneous fission, a second dataset is required.

Secondary particles access. The ENSDF database primarily focuses on nuclear structure data while also incorporating reaction data, including decay data. Its evaluation methodology examines entire mass chains simultaneously, resulting in comprehensive data, see Table 2. This extensive approach motivated its adoption for the GENF/D project, as it avoids a supplementary cross-compilation of additional datasets, as the ones from DDEP.

However, as noted previously, parsing ENSDF files presents challenges due to their experiment-centric organisation. Critical information can be embedded within comments, and secondary particle radiation data are distributed across multiple sections, complicating automated extraction. To address these parsing difficulties, the Livechart API offers an alternative access route. This interface provides restructured ENSDF data with simplified access, along with supplementary decay-related information when required.

While the Livechart web interface provides supplementary data including estimated beta spectrum energy distributions and internal bremsstrahlung, these have limitations. The beta spectra derive from an outdated BetaShape version, while the bremsstrahlung data (from RadiationReport) remains inaccessible via the API. For GENF/D building, we regenerated both spectra independently (see details below).

Although Livechart includes total conversion coefficients for the description of conversion electrons, details on K-shell and L-shell internal are missing for the associated photon spectrum. We employed the BrIcc code to calculate these values, enabling the inclusion of RICK and RICL parameters of ENDF-6 format in GENF/D.

Finally, to achieve comprehensive self-consistency (particularly for excited level coherence and complete fragment lists for all relevant nuclides) we generated a spontaneous fission yield library using the GEF code. This implementation provides Q-values and average secondary energies for spontaneous fission processes.

The following sub-sections detail the specific codes and methodologies employed in this work.

Beta and neutrino spectra. For decades, the estimation of beta spectra for nuclear data evaluations has primarily relied on the LogFT code. Although capable of handling beta and electron capture transitions and propagating uncertainties from input parameters, LogFT used simple analytical models to ensure fast calculations, which led to a lack of precision, particularly in screening effect corrections and shape factors for forbidden transitions.

To overcome these limitations, the BetaShape code was developed. BetaShape brings improved physical models, a database of experimental shape factors, and provides more detailed information such as beta and neutrino spectra. BetaShape is a code developed by the Laboratoire National Henri Becquerel (LNHB/CEA) to compute beta decays with the goal to improve theoretical predictions of beta decays, electron capture, neutrinos and capture probabilities. The programme also calculates beta and neutrino energies, together with log-ft values for each transition. It includes an analytical screening correction, which leads to satisfactory results above a few keV, but lacks of precision at very low energies for an accurate description of the beta spectrum. From version 2.3, more accurate atomic effects are included (via extensive tabulation of numerical solving): screening, exchange, overlap [32]. BetaShape version 2.2 was used by the Livechart API maintainers, using default parameters. In this work, the version 2.4 has been used together with the ENSDF data as of March 2025, as present in the IAEA Livechart. The default parameters have also been systematically used.

Internal conversion factors. BrIcc, distributed by the Australia National University (ANU), allows the computation of theoretical coefficients for internal conversion electrons (ICC), electron-positron pair conversion coefficients (IPC), and electronic factors Ω(E0) for nuclear transitions of pure or mixed multipolarity. BrIcc leverages state-of-the-art theoretical models to provide reliable calculations of these coefficients.

The code employs a self-consistent relativistic Dirac-Fock (DF) approach, representing a significant advancement in theoretical precision, with results now rivalling experimental accuracy at the percent level. Comparisons with experimental data demonstrate superior agreement with BrIcc’s DF-based calculations compared to older tabulations such as HsIcc (Hager and Seltzer) and RpIcc (Rösel et al.). The default implementation, BrIccFO, incorporates the frozen orbital approximation to account for atomic vacancies during conversion processes, showing better alignment with experiments than the no-hole approximation (BrIccNH) [29]. Adopted internationally by the NSDD network, BrIcc is now the standard tool for new evaluations in Nuclear Data Sheets and Nuclear Physics A [33]. It spans broad transition energy ranges (1–6000 keV for ICC; 1100–8000 keV for IPC) and atomic numbers (Z = 5–110), exceeding the scope of earlier tabulations. Additionally, BrIcc systematically evaluates conversion coefficient uncertainties by incorporating theoretical errors, transition energy uncertainties, and mixing ratio uncertainties.

The version 2.3d of BrIcc slave program has been ran, with default parameters (e.g. Frozen Orbital approximation, or using Tab. 3 for missing mixing ratios), using the gamma information available in the Livechart API. Compared to the total coefficients already available through the Livechart, some deviations have been identified but none greater than 4%.

Internal bremsstrahlung. The RadiationReport Java programme is among the code recommended by the NSDD network and part of the ENSDF Analysis and Utility Programs, as BetaShape and BrIcc. It allows the calculation of energies, intensities and doses of all radiations as well as log-ft values for a decay dataset. It is an alternative to the RADLIST and LogFT Fortran codes combined. This code has been run in order to obtain the internal bremsstrahlung continuous spectrum not available through the Livechart API.

The version released in April 2025 have been used in this work, and as for BetaShape rerun, the ENSDF files as of March 2025 were used as inputs. The default parameters have been systematically used.

Spontaneous fission. The GEF model (GEneral description of Fission observables) is designed to deliver a comprehensive and self-consistent description of fission observables for spontaneous fission, neutron-induced fission, and compound-nucleus fission across arbitrary entrance channels, over a wide range of excitation energies and angular momentum states. Designed to supply nuclear data for reactor technology and engineering applications, GEF covers an extensive isotopic range (Z = 80 to 112 and beyond) and excitation energies up to ∼100 MeV, including multi-chance fission processes.

The GEF code employs a Monte-Carlo framework to simulate fission dynamics while preserving correlations between observables, offering critical insights into the fission process. Unlike empirical models that directly parametrise observables, GEF integrates approximations rooted in microscopic system properties, with parameters globally optimised to minimise deviations from experimental data. This physics-based approach ensures robust predictive power for nuclei near experimentally characterised cases, though localised structural effects may introduce exceptions. The model successfully reproduces key features such as delayed neutron probabilities – linked to fragment yields and excitation energies – and provides detailed predictions for total kinetic energy (TKE) distributions as functions of fragment mass. By combining general physical principles with empirical constraints, GEF achieves superior predictive accuracy and evaluative capability compared to alternative methods, enabling reliable extrapolations for unmeasured fission quantities across a broad range of nuclei.

One should note that, GEF does not include ternary fission.

This code was chosen over other codes, such as FREYA [34] or FIFRELIN [35], as it allows to directly obtained an ENDF-6 formatted output for fission fragment list, as well as access to the secondary spectra (neutron, photons, fission fragments energy). Further motivations are available in the validation Section 4.1.2.

The latest version (2025.1.2) has been used for this project. In addition, a nuclear property table and branching table have been generated based on GENF/D to ensure consistency in the excited states, and for the cumulative yields generation.

3. Generation of the decay database

A pythonic toolbox has been developed to handle the relevant data.

The environment is built on the IAEA’s endf_parserpy package [36], together with pandas [37] and pint (for uncertainty propagation) packages. One should note that as the targeted usage of the GENF/D library is for codes such as DCHAIN-PHITS or TMX_Bateman, minimal handle of uncertainties was done. The uncertainties are propagated using Taylor expansion, and when not present in the original media, the uncertainty attached to any quantity was set to 100%.

The ENSDF data are accessed via the IAEA Livechart API, except for use of BetaShape and RadiationReport, where it is downloaded from the National Nuclear Data Center. The NUBASE data is retrieved directly from the Atomic Mass Data Center website.

These datasets are processed in several stages to produce a full and valid ENDF-6 formatted decay sub-library:

• Step 1:

  conversion of NUBASE2020 to an ENDF-6 decay library, supplemented with RIPL-3.23 spin estimates.

• Step 2:

  implementation of decay spectra from ENSDF, accessed via the IAEA LiveChart API, supplemented with BetaShape and BrIcc estimates.

• Step 3:

  implementation of Spontaneous Fission (SF) Q-values and associated neutron and photon spectra from the GEF code.

Figure 1 illustrates these steps in a general flowchart.

Fig. 1. Flowchart of data source for GENF/D generation.

3.1. Step 1: generation of basic ENDF-6 database

The basic data is first extracted from the NUBASE2020 and AME2020 databases. A radioisotope, and its channels, are retained if the decay modes and branching data are available, and if the Q-value is positive (taking the excitation energy into account).

The processing of the decay modes and their branching ratios are described in Section 3.1.1. In this section, the filtering choices are described in detail.

If an isotope has no half-life provided, it is discarded. This happens for 63 radioisotopes (with 46 ground states). The exhaustive list is ¹²Li, ²³C, ¹³F, ¹⁷Na, ⁴¹Mg, ²⁴P, ³²K, ^32mK, ³³Ca, ³⁵Sc, ³⁶Sc, ³⁷Sc, ^38mSc, ³⁹Sc, ³⁷Ti, ³⁹V, ⁴⁰V, ⁴¹V, ⁴⁴ⁿV, ⁴¹Cr, ⁴³Mn, ^47mFe, ⁴⁷Co, ⁴⁸Co, ⁵²Cu, ⁸⁴Cu, ⁸⁶Zn, ⁵⁶Ga, ⁵⁷Ga, ⁵⁸Ga, ^58mGa, ^61mGa, ⁸⁸Ga, ⁵⁸Ge, ⁶⁰As, ^60mAs, ⁶¹As, ⁶²As, ⁶⁷Br, ^69mBr, ⁶⁹ⁿBr, ⁷¹Rb, ^71mRb, ⁷¹ⁿRb, ^72mRb, ^73mRb, ⁷⁹Nb, ⁸⁰Nb, ⁸³Tc, ⁸⁴Tc, ¹³³Ag, ¹³⁵Cd, ¹⁰²Sb, ¹⁰⁶I, ¹⁵⁹Ce, ¹⁷⁰Eu, ^145mTb, ¹⁸²Tm, ¹⁶³Ir, ¹⁶⁸Au, ^178qAu, ^192pTl, ^250pBk. The majority of these radioisotopes, being close to the drip lines, are subjected to quite different physics regimes, making an estimate of their half-life impossible.

It should be noted that energies can be negative due to the experimental uncertainties. In such a case, the RIPL methodology is applied, i.e. shifting the negative energy by its uncertainty, or alternatively flooring it to zero (if the former attempt was not successful). This happens for 9 excited isotopes, explicitly for: ^91mRu, ^124mIn, ^154mPm, ^152mTm, ^164mRe, ^194mAt, ^196mAt, ^248mBk, ^265mSg. As an example, the excitation energy of ^91mRu reported to be −340 ± 500 keV, is shifted to 160 keV.

The same process is applied to the Q-values of 13 isotopes: ³¹Ar, ⁵⁵Ti, ⁶⁹Fe, ⁷⁸Ni, ⁶⁸Kr, ⁸⁵Ru, ¹⁷³Dy, ¹⁷⁹Tm, ¹⁸⁶Lu, ²¹⁹Bi, ²²⁵At, ²⁶⁰Rf, ²⁹⁰Fl. For example, the Q_β⁻n of ⁵⁵Ti given in NUBASE2020 is −6 ± 31 keV: this channel is raised to 25 keV. Flooring to zero was only necessary for ^154mPm.

Though present in NUBASE2020, the spins and parities included in GENF/D are the one available in RIPL-3.23. Indeed, some states in NUBASE can exhibit several spins due to incomplete experimental knowledge. However, the ENDF-6 format restricts spin to a single value. To refine spin assignments, the RIPL project uses experimental spin distributions and statistical models [38, 39]. GENF/D prioritises these estimates.

Nevertheless, RIPL does not provide spin assignments for 621 isotopes. For these cases, NUBASE spins are used as a fallback. If NUBASE lists multiple spins (e.g., ¹⁰Li with (1⁻, 2⁻)), the first value (1⁻) is selected. The situation arises for 64 isotopes. Finally, 250 isotopes have unknown spins.

In addition, Q value are computed, see Section 3.1.2, and the average energy of the products is estimated, see Section 3.4.

This procedure produces the minimum required information to track a radionuclide inventory. The description of each file (MF = 1/MT = 451 in ENDF-6 format) gives the raw information from NUBASE in case of doubt on data.

3.1.1. Conversion of NUBASE decay to ENDF-6 format

This section details the conversion of the decay mode and branching from NUBASE, as several non straight-forward choices were made.

The ENDF-6 format encodes the decay into a real number, called RTYP. Each emitted particle is associated with a digit, see Table 4. Multiple particle decay is encoded by adding significant digits e.g. a beta decay followed by an alpha emission is registered as 1.4. The delayed neutron emissions can be described in the same manner, e.g. marking 1.55 for a β⁻ emission followed by two delayed neutrons. In addition, ENDF-6 format does not distinguish positron emission and electron capture (except when the beta spectrum is given), so the splitting indicated by NUBASE is ignored.

Yet NUBASE presents decay modes that cannot be encoded directly into RTYP, such as “cluster decays”, i.e. emission of ¹²C, ¹⁴C, ¹⁸O, ²⁰Ne, ²⁰O, ²²Ne, ²³F, ²⁴Ne, ²⁴Ne+²⁶Ne, ²⁵Ne, ²⁸Mg, ²⁸Mg+³⁰Mg, ³⁰Mg, ³²Si, ³⁴Si. Amongst all, only the ¹²C emission is kept (with a maximum branching appearance of 0.0034%) and translated using RTYP = 4.44, meaning a 3α decay. The others cannot be described as easily but having a branching ratio less than 8.9 × 10⁻⁸%, they can be safely ignored.

In addition, the deuterium and tritium emissions are converted in the 2p and 3pRTYP equivalents. The concerned channels are the ${\beta }^{-}{+}_1^2$H of ⁶He and ¹¹Li and the ${\beta }^{-}{+}_1^3$H of ⁶He, ¹¹Li and ¹⁴Be. This does not overlap with any β⁻ + 2p and β⁻ + 3p channels, as they are not part of recognised decay channels.

The ENDF-6 format requires that the total branching ratio sums up to 100%, at variance with NUBASE format. For example, for the decay of ¹¹Li, in NUBASE format, the decay channels are given as 100% emission of β⁺ and 86% of β⁺n emission, whereas ENDF-6 indicates 14% of β⁺ and 86% of β⁺n. This typical case happens for 529 isotopes. After all such corrections and in any case, the sum of the partial branching ratios is systematically renormalised to 1.

This entails to know every branching for all accounted channels, which is not the case. Discarding the isotopes when no half-life is provided, over the 5248 unstable isotopes available, 1519 have at least one channel without branching. The main actions taken are categorised in Table 5, the last column of this table giving one example for illustration. The different corrections are listed by priority. There are actually 793 nuclei that need renormalisation to obtain 100%. After the prompt emission was corrected, negative Q-value channel, and unknown decays are removed, and the positron/electron-capture splitting is merged.

3.1.2. Available decay energy

In order to estimate accurate heat production, precise computation of Q-value is necessary. This database computes the total available decay energy based on the atomic masses of AME2020 database. The total available energy, for a decay channel, considering the conservation of energy E, is as follows:

$$\begin{array}{cc}\hfill Q& =E(X_i)-\sum _{f}E(Y_f)\hfill \\ \hfill & =M_{\mathit{atm}}(X_i)c^2+E^{\ast }(X_i)-\sum _f(M_{\mathit{atm}}(Y_f)c^2+E^{\ast }(Y_f))\hfill \end{array}$$(1)

where E^* is the excitation energy, X_i the parent nuclide, Y_f the daughter products, and M_atm is the atomic mass defined by:

$$\begin{array}{c}\hfill M_{\mathit{atm}}{(}_Z^{A}X)=M_N{(}_Z^{A}X)+Z\times m_e-B_e(Z)/c^2\end{array}$$(2)

where A is the atomic mass number, Z the atomic number and c the speed of light in vacuum. An approximation of electron binding energy, in electron-volt, is given by the empiric formula [10]:

$$\begin{array}{c}\hfill B_e(Z)=14.4381\times Z^{2.39}+1.55468\times {10}^{-8}\times Z^{5.35}.\end{array}$$(3)

There are 29 decay modes, described by the RTYP parameter, available in the generated sub-library, which are β⁻, 2β⁻, β⁻ + α, β⁻ + n, β⁻+2n, β⁻+3n, β⁻+4n, β⁻+SF, β⁻+p, β⁻+2p, β⁻+3p, β⁺, 2β⁺, β⁺ + α, β⁺+SF, β⁺ + p, β⁺+p+α, β⁺+2p, β⁺+3p, IT, α, 3α, n, 2n, 3n, SF, p, 2p, 3p. The electron capture distinction for the β⁺ (and extended) are the responsibility of the end-users.

As there is almost no mention of isomeric states of daughters, most Q-values computed in GENF/D consider the daughter at ground states, i.e E^*(Y_f) is null in equation 2. The eleven exceptions are ²⁴Ne, ⁷⁷Ga, ⁹⁰Nb, ¹⁰⁹ⁿIn, ¹¹⁴ⁿIn, ¹¹⁶ⁿIn, ¹¹⁸ⁿIn, ¹²⁴ⁿSb, ¹²⁶ⁿSb, ¹⁴⁸ⁿHo and ¹⁵⁰ⁿTn.

The available energy for all decays in GENF/D are computed based on the raw definition of decay (and not using the RTYP alternative). Following the convention of other evaluated databases, $E_{{\beta }^{+}}/c^2=Q_{\mathit{ec}}/c^2=M_{\mathit{atm}}\left({}_Z^{A}X\right)-M_{\mathit{atm}}\left({}_{Z-1}^{A}Y\right)$.

The values of Q_β⁻ and Q_α, available directly in NUBASE, are preferred by GENF/D, but corrected with parent and daughter excitation energy. Below listed all Q-values recomputed (not accounting for excitation levels of parent and daughter nuclides) are:

$$\begin{array}{r}-Q_{\beta -\mathrm{xn}}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z+1}^4\mathrm{Y}\left|\right)-x\times {\mathrm{mn}}_{}\\ -Q_{\beta -\mathrm{xp}}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z+1}^4\mathrm{Y}\left|\right)-x\times \\ M_{\text{atm }}\left(\right|{}_1^{2}H\left|\right)\\ -Q_{\beta -2}\mathrm{H}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_Z^{4-2}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_1^{2}H\left|\right)\\ -Q_{\beta -3}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_Z^{4-3}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_1^{4}H\left|\right)\\ -Q_{\beta -\alpha }/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-1}^{4-4}\mathrm{Y}\left|\right)-M_{\text{atm }}\left(\right|{}_2^{4}He\left|\right)\\ -Q_{{\beta }^{+}}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-1}^4\mathrm{Y}\left|\right)\\ -Q_{2\beta +}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-2}^4\mathrm{Y}\left|\right)\\ -Q_{{\beta }^{+}\mathrm{xp}}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-1}^4\mathrm{Y}\left|\right)-x\times \\ M_{\text{atm }}\left(\right|{}_1^1\mathrm{H}\left|\right)\\ -Q_{{\beta }^{+\alpha }a}/c^2=M_{\text{atm }}\left(\right|{}_Z^4\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-3}^{4-4}\mathrm{Y}\left|\right)-M_{\text{atm }}\left(\right|{}_2^4\mathrm{He}\left|\right)\\ -Q_{\beta +\mathrm{p\alpha }}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-4}^{A-5}\mathrm{Y}\left|\right)-\\ M_{\text{atm }}\left(\right|{}_2^4\mathrm{He}\left|\right)-M_{\text{atm }}\left(\right|{}_1^{1}H\left|\right)\\ -Q_{\mathrm{xn}}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_Z^{A-x}\mathrm{X}\left|\right)-x\times {\mathrm{mn}}_{}\\ -Q_{\mathrm{xp}}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-x}^{A-x}\mathrm{Y}\left|\right)-x\times M_{\text{atm }}\left(\right|{}_1^1\mathrm{H}\left|\right)\\ -Q_{2_{\mathrm{H}}}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-1}^{A-2}\mathrm{Y}\left|\right)-M_{\text{atm }}\left(\right|{}_1^2\mathrm{H}\left|\right)\\ -Q_{3_{\mathrm{H}}}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-1}^{A-2}\mathrm{Y}\left|\right)-M_{\text{atm }}\left(\right|{}_1^3\mathrm{H}\left|\right)\\ -Q_{12}\mathrm{C}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-M_{\text{atm }}\left(\right|{}_{Z-6}^{A-12}\mathrm{Y}\left|\right)-M_{\text{atm }}\left(\right|{}_6^{12}\mathrm{C}\left|\right)\\ -Q_{\mathrm{SF}}/c^2=M_{\text{atm }}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)-\sum {\mathrm{yi}}_{}\times M_{\text{atm }}\left(\right|{\mathrm{Y}}_i\left|\right),\text{ with }\\ \sum _i{\mathrm{yi}}_{}=2\\ -Q_{\mathrm{IT}}=E^{*}\left(\right|{}_{\mathrm{ZA}}^{}\mathrm{X}\left|\right)\end{array}$$

In Figure 2, we summarise the main decay modes for the ground state of the nuclei available in our library.

Fig. 2. Segrè diagram of available nuclides and their main decay modes in the GENF/D decay sub-library.

3.2. Step 2: insertion of decay spectra

This step is particularly important for decay heat applications, which require precise spectral data for all emitted particles to estimate the average energy accurately, in order to be able to access each decay heat component. The decay radiation characteristics (i.e. ray energies and intensities) are extracted from ENSDF, thanks to the LiveChart API, and completed with BetaShape (beta and neutrino continuous spectra) and BrIcc (internal conversion coefficients for gamma rays) estimates where relevant.

Livechart provides the α, β, γ, X-rays and discrete electrons (conversion and Auger). The main source is the ENSDF archive as of March 2025, with the atomic shell transition yields from [40] and atomic shell energies from the Evaluated Atomic Data Libraries EPICS2023 [41].

Delayed spectra are not available via the API, but some of the data was alternatively made available in [42, 43]. The delayed-neutron spectra of ¹⁴Be, ¹⁵B, ¹⁷B, ¹⁴¹Cs, ¹⁴²Cs, ¹⁴³Cs, ¹⁴⁴Cs, ¹⁴⁵Cs, ¹⁴⁶Cs, ¹⁴⁷Cs, ¹⁶C, ⁸He, ¹²⁹In, ¹³⁰In, ¹³⁷I, ¹³⁸I, ¹³⁹I, ¹⁴⁰I, ¹⁴¹I, ⁹Li, ¹¹Li, ²⁷Na, ²⁸Na, ¹⁷N, ¹⁸N, ²¹N, ¹³⁵Sb, ¹³⁴Sn, ¹³⁶Te, ⁷⁹Ga, ⁸⁰Ga, ⁸¹Ga, ⁸⁵As, ⁸⁷Br, ⁸⁸Br, ⁸⁹Br, ⁹⁰Br, ⁹¹Br, ⁹²Rb, ⁹³Rb, ⁹⁴Rb, ⁹⁵Rb, ⁹⁶Rb, ⁹⁷Rb have been included in GENF/D, as histograms to the neutron continuous spectrum (STYP = 5).

The obtained Livechart API spectral information is supplemented with BetaShape (version 2.4) and RadiationReport (version of April 2025), run with ENSDF2025 (March 2025), to include improved β, capture and neutrino spectra thanks to the former, and internal bremsstrahlung using the latter code. When available, the BetaShape information fully replaces the API information. The BrIcc programme has been run to generate again the internal conversion coefficient, in order to access the K- and L-shell coefficients (RICK and RICL), based on gamma ray information provided by the Livechart. Some deviation with the total conversion coefficient, which was estimated by Livechart maintainers also with BrIcc, have been identified but none greater than 4%.

Whenever a spectrum is available, it is appended to the skeleton file. The previous versions of ENDF-6 format, as [44], described in the β spectra section both the charged particle (e⁺, e⁻) and the neutrino (or anti-neutrino). According to the 2018 update of the format [45], the neutrino part can be removed from the spectra related to beta decay and placed in the dedicated section (STYP = 10 or 11). This distinction has been applied within GENF/D. In addition, when positrons are emitted, the two photons due to annihilation are added to the x-ray spectrum (STYP = 8) as recommended by the ENDF-6 format.

All mentioned spectra are added as discrete spectra, except for the BetaShape ones, the internal bremsstrahlung and the neutron delayed spectra. The latter is a continuous spectrum, that contains the positron-to-beta ratio in its integral rather than in the RIS entry, when given as discrete rays as was done until this library.

3.3. Step 3: insertion of spontaneous fission properties

In the last step, we compute the Q-value of spontaneous fission (SF), as well as the spectra of neutrons, photons and fission fragments. Only a few experimental data is available in ENDF-6 format, with only ten evaluated spontaneous fission fragment lists released altogether in ENDF/B, JEFF and JENDL.

The fission fragment lists have been generated using the Monte Carlo GEF code, version 2025.1.2, with 10⁹ events simulated, based on the property tables extracted from our GENF/D library, for consistency. The subsequent sub-libraries for spontaneous fission yields (SFY) and neutron-induced fission yields (NFY) were produced, thanks to the GEF features. These data will be released alongside the GENF/D library, and are used in the benchmarks, discussed in Section 4.2.

The fission fragment list, as well as neutrons and total kinetic energy of fragments spectra have been produced for 338 isotopes (including 279 ground states). The neutron and kinetic energy of fission fragments spectra have been imported in GENF/D as continuous spectra. Obviously, the average neutron multiplicity is also added in the relevant section of the decay file.

3.4. Average energy estimations

ENDF-6 format allows to indicate the global average energy of the emitted particles. One can list three main decay heat components, i.e. light particles (β⁻, β⁺, electrons, …), electromagnetic radiations (γ, x-ray, annihilation radiation, …) and heavy particles (α, FF, p, n, …) average energies.

The format allows to describe the decay heat component of 14 channels, belonging to the three categories mentioned above. Such feature is not used in GENF/D. Indeed, NUBASE does not provide detailed information about the nucleus relaxation (intermediate levels in daughter nucleus, gamma or particle emissions). This prevents any realistic determination of the splitting of the decay energy in the different heat components, without the ENSDF data.

Through the IAEA API, without considering the different isomers, there are 2158 isotopes (to compare to the 3511 ground states of GENF/D) with information on the associated gamma transitions and 813 with β⁻, 797 with β⁺, and 683 with α information. Therefore, even with ENSDF data, all average energies of particle emitted cannot be computed, and estimates must be determined.

From a detailed analysis of ENDF/B heat estimation when no spectra were given, we extracted a convention in order to get the average energies from the total available energy. These rules are presented in Table 6. Figure 3 shows the relative difference between original ENDF/B and ENDF/B average energies recomputed with Table 6 convention, for the nuclides without any decay spectra information, considering 10²³ W total heat for each radionuclide.

Fig. 3. Log-scaled relative difference between ENDF/B original and recomputed from Q-value decay heat by atomic number Z. The discrepancies for the total decay heat (blue hexagon), as well as light particle (yellow diamond), electromagnetic (green square) heavy particle (grey bullet) decay heat are represented, with their global average.

For those 1828 nuclides, the average relative error is 0.005 ± 0.05%. The decay heat components from each radionuclide were compared on a normalised basis, assuming a total decay heat output of 10²³ W per nuclide.

In ENDF/B, JENDL, and JEFF, the heat released from spontaneous fission, as well as the heat resulting from β-decay followed by spontaneous fission, is typically assigned a null Q-value. Except for the ten nuclides that do have ENDF-6 translated fission fragment list (²³⁸U, ²⁴²Cm, ²⁴⁴Cm, ²⁴⁶Cm, ²⁴⁸Cm, ²⁵⁰Cf, ²⁵²Cf, ²⁵³Es, ²⁵⁴Fm, ²⁵⁶Fm). As GEF can provide the list of fragments, and their combined kinetic spectra, as well photo-fission spectra, there is no need to find any estimate based on the reaction Q values.

The fraction of Q-value is then scaled by the branching ratio of each decay channel, in order to get the heat component per decay process.

When spectra are available, the associated average decay heat component is instead determined with equation 5.

$$\begin{array}{c}\hfill ⟨E⟩=BR\times \frac{\int P(E)EdE}{\int P(E)dE}\end{array}$$(4)

where BR is the branching ratio of the considered channel, and P is the emission probability at energy E.

For the 5502 isomers, only 1962 kept the guess estimate for all three values.

4. Discussions

To summarise, the GENF/D sub-library provides information on 5502 radioisotopes, including 1990 isomers, with 8531 decay channels split in 28 unique decay modes (unique RTYP). In the section below, the sub-library is validated against the other main databases and in a second time, against heat benchmarks.

4.1. Comparison of GENF/D against other databases

The primary design objective for the GENF/D library is extensiveness. A comparison with all other publicly available ENDF-6 format libraries reveals that the combined set contains data for 5583 isotopes (comprising 3567 ground states and 2016 isomers). The GENF/D library is very complete but misses 49 ground states and 33 isomeric states present in other libraries.

The majority of these omitted nuclides are meta-stable with very short half-lives falling below the T_1/2 ≥ 100 ns validity threshold of the NUBASE2020 evaluation (e.g., ¹²Li, ⁵⁶Ga). As for some of the longer-lived nuclides, they simply do not appear in NUBASE2020, i.e. ²⁴⁴U, ²⁴⁵U, ²⁴⁶Np, and ²⁵⁰Am, nor do they appear in ENSDF, making difficult their addition to GENF/D in a direct manner.

A potential solution to these omissions would be to import data for the missing isotopes directly from other evaluated databases. This approach, however, presents significant challenges. Data for some nuclides is only available in outdated distributions (e.g., 5 in ENDF/B-VII.0, 3 in JEFF-2.2, and 1 in JENDL-FPD-2000) and was not carried forward into subsequent updates of those libraries, raising doubts about its reliability.

Furthermore, while more credible modern evaluations – such as JENDL-5 (59 isotopes), JEFF-4.0 (9 isotopes), and EAF-2010 (4 isotopes) – could potentially fill most of the gaps, a conscious decision was made to exclude this data. The priority was to ensure absolute internal consistency within the decay chains and to maintain a uniform methodology for spectral association across the entire library.

4.1.1. Energetics considerations

As a reminder, GENF/D proposal is the translation of the evaluated NUBASE in ENDF-6 format, and compiled with the evaluated ENSDF. Hence, some discrepancies are expected against the other decay sub-libraries as ENDF/B, JEFF or JENDL as they do a second layer of evaluation. Validation of GENF/D can be quite complex due to the high number of nuclides. In this section we provide global check against ENDF/B, JEFF and JENDL. To keep the reference consistent in the relative differences, we have chosen to use GENF/D as the most extensive one.

Firstly, in order to check the correct parsing of our half-lives, the ones of GENF/D have been compared in Figure 4. As expected, as they are closer in release date and sources, GENF/D half-lives are in better concordance JENDL-5, with 80% of half-lives within 5% agreement. The other main properties have been compared and synthesized in Table 7.

Fig. 4. Relative difference between GENF/D half-lives and JENDL-5.

Another relevant and easily comparable property is the total available decay energy. The agreement within 5% is of 77% for JEFF-4.0, 82% with ENDF/B-VIII.0 and 81% with JENDL-5. If we consider the maximum decay energy, then the agreement is over 82% for all of them.

As seen in Table 7, the decay heat estimates have lesser agreement, especially with JENDL-5. This is mainly due to mistakes in the decay heat estimation of JENDL-5. Indeed, some decays that emit delayed neutrons do not account them in the heavy particle heat.

As for the consideration of delayed neutrons, one should note that, without accounting for fission neutrons, JEFF-3.3 have neutron average decay energy for 80 isotopes, ENDF/B-VIII.1 for 312 isotopes and JENDL-5 for 494 isotopes, whereas GENF/D have data for 44 isotopes. This reduced number of neutron spectra in GENF/D is due to the use of the IAEA compilation, rather than extracting the one from ENSDF, as they are not available through the Livechart API.

4.1.2. Spontaneous fission considerations

In this section, we present the comparison of JEFF-4.0 with our results using GEF, to confirm the motivation of its use, and validate the GENF/D estimates. The fission data is virtually the same for JEFF, ENDF/B and JENDL when considering the decay data, except for ²³¹Pa that is absent in JENDL-5. Since JEFF-4.0 provides more details on the associated spectra, we chose this database for comparison.

JEFF indicates 132 channels of SF or β followed by fission, and provides non-null Q value for only 27 of them. In contrast, GENF/D indicates 240 isotopes that undergo spontaneous fission, and provides an estimate of Q-value for all of them, as well as neutron-multiplicity and the neutron, gamma and fission fragment spectra.

When comparing the 27 isotopes, which have detailed data in JEFF-3.3, the available decay energy difference between GENF/D and GEF is within 2%. Hence, even if GEF does not consider the ternary fission, because the probability of appearance of this phenomenon is very reduced, there is no impact on energy estimations.

If we compare the neutron multiplicity, meaning the average number of neutrons emitted after a spontaneous fission, the discrepancies get larger with a maximum of 21%, and an average discrepancy of 7%. This remains in an acceptable range of difference, which are shown in Figure 5.

Fig. 5. Relative discrepancy between GENF/D and JEFF-4.0 for available decay energy (left) and neutron multiplicity (right).

Table 8 illustrates this comparison for ²³⁸U and ²³²Th decays. For these isotopes, the average neutron energies are in perfect agreement, within at most 1.2% of difference. The gamma spectrum disagrees by only 2% for ²³⁸U, but jumps to 14% for the ²³²Th. The difference in the gamma spectrum with JEFF-4.0 could come from the fact that the relaxation spectra of such a decay are represented by a very smoothed continuous linear-linear distribution, whereas GEF generates a histogram with 1 keV binning. It should also be noted that ternary fission is not considered in GEF, leading to the absence of light fragments as protons, helium, lithium, etc. that can be found in the other evaluations.

Globally, the average energy of fission fragments have in average 0.9% relative difference, for a maximum discrepancy of 3.9% for ²⁵³Es, with respect to JEFF-4.0. Due to the higher discrepancy with the average neutron number emitted, the neutron average energy has a larger discrepancy, in average of 11%.

4.1.3. Average energy from Q-value

When no experimental spectra is available, the convention for estimating average energy from available energy (Tab. 6) is applied. Yet, there are some concerns, that should be known.

When applying such convention, to ENDF/B-VIII.0 decay sub-library, on nuclides that do have information on secondary particle energy distributions, one observes significant discrepancies (Fig. 6). For these nuclides, having at least one information on α or β secondaries in ENDF/B-VIII.0, the average total decay energy difference is 96.4 ± 747% (excluding deviations larger than 1000%).

Fig. 6. Log-scaled relative difference between ENDF/B decay heat estimates (considering 1023 W total heat of each radionuclides) and the one recomputed based on Table 6 convention, considering only the nuclides with spectral information. The discrepancies for the total decay heat (blue hexagon), as well as light particle (yellow diamond), electromagnetic (green square) heavy particle (grey bullet) decay heats are represented, with their global average.

The electromagnetic decay heat component is particularly affected, with an average discrepancy of about 10⁶% due to systematic overestimation. The heavy particle component is less affected, with an average discrepancy of 28%.

Since the JEFF and JENDL libraries use a comparable convention (showing discrepancies of 3.7 ± 10.5% and 0.04 ± 0.8%, respectively), the agreement in Table 7 is not an independent validation of the convention’s correctness but an indication of its consistent application.

The assumption that light particle decay heat equals one-third of the available energy is arguably arbitrary. This is shown in Figure 7, which plots the ratio of the average beta energy, calculated using BetaShape, to the corresponding Q-value from NUBASE2020. Empirical data suggest more accurate mean values of approximately 0.26 for β⁺ emissions and 0.31 for β⁻ emissions. But finer models, accounting for spins, should bring better predictions.

Fig. 7. Ratio of the average beta energy (from BetaShape) to the NUBASE2020 Q-value. The top panel shows results for β+ emissions, and the bottom panel for β− emissions. Data are shown for distinct isomeric states (I = 1 corresponds to first isomeric states, in darker tones, and I = 0 to ground states, in lighter tones), alongside the global average and a linear fit over atomic number Z.

4.2. Validation against integral measurements

A critical aspect of the development of a decay database is its validity against decay heat benchmarks. Only a few of them are available due to the difficulty of such measurements.

Due to the decay heat estimation strategy, we consider only total decay heat comparison in this section, in order to avoid pandemonium effect and the questionable decay heat splitting.

It should be mentioned, that the pandemonium effect is not considered in GENF/D. Introducing the Total Absorption Gamma-ray Spectroscopy measurements, in order to access the higher energy beta transitions causing the pandemonium effect [46], and consequently reevaluate all derived transitions is beyond the scope of this first release. If such correction is critical for end-user application, we recommend to use the evaluated datasets provided by projects like JEFF [13], that does partially introduce such considerations, e.g. for ⁹³Rb, ⁹⁶Y, ^96mY, ⁹⁹Y, ¹⁰³Tc, ¹⁰⁸Tc, ¹³⁸I or ¹⁴²Cs.

To produce the results presented in this section, we have used FISPACT-II. This code follows the evolution of materials irradiated by neutrons, alphas, gammas, protons, or deuterons, and provides a wide range of derived radiological output quantities to satisfy most needs for nuclear applications.

In Figure 8, we present two transuranic decay heat benchmarks, the results of 10^-6 seconds of thermal pulse (with a fluence of 10²² n/s) on pure ²³⁹Pu, estimated by Dickens [47] and Tobias [48], and the results of 2 × 10⁴ seconds of irradiation of pure ²³⁵U (with a fluence of 10¹³ n/s) by Yarnell [49].

Fig. 8. Decay heat benchmark using ENDF/B-VIII.0 neutron cross-section for GENF/D against ENDF/B, JEFF and JENDL decay data (using their associated fission fragments). On the left, a long thermal irradiation on 235U. On the right, a thermal pulse experiment on 239Pu.

In Figure 9, we present an irradiation of non fissile material, with an irradiation of alloy (Inconel 600) for 5 min with a fluence of 1.116 × 10¹⁰ n/s at the Fusion Neutron Source (FNS).

Fig. 9. Decay heat benchmark using TENDL-2017 neutron cross-section, GEFY fission fragments and GENF/D against ENDF/B, JEFF and JENDL decay data. Five minutes irradiation at FNS of natNi, natMn, natFe, natCr.

One should remember, that such benchmarks do not depend only on the decay database, but also heavily on the fission fragment estimates, in the case of fissile material, and on the neutron cross-sections for the FNS irradiation.

Figures 8 and 9 were generated using GEFY induced and spontaneous fragments. Due to the heavy link between GEF and GENF/D, we recommend the use of GEFY with GENF/D. But no correction (or evaluation) of the data was done to meet the experimental points, and create more links than already mentioned in Section 3.3.

5. Conclusion

We have produced a novel decay database from NUBASE2020 and ENSDF2025 data in ENDF-6 format. It accounts for the new updated format rules, to discriminate neutrino spectra from their beta counterpart.

Summary of contributions: the main objective of GENF/D was to produce a dataset extending far beyond the centre of the stability valley, including exotic and meta-stable nuclides. This was achieved by translating the NUBASE datasets into valid ENDF-6 format, thanks to the endf_parserpy utility, as well as IAEA LiveChart API.

The methodology and measurements have been confirmed over the common nuclides with the other evaluated database. Additional excited state data have been made available (from 797 excited states in JENDL-5, 768 in JEFF-3.3 or 738 in ENDF/B-VIII.0 to 1990 in GENF/D), enabling more comprehensive follow of inventory, especially in the case of induced and spontaneous fissions that both produces a large number of excited states.

This work also led to the production of a more complete GEFY fission fragment dataset. In addition, we have highlighted some errors in the other database (as missing rays, contribution to heat, …), that we transmitted to their maintainers.

The release of this new decay database marks a significant step forward, with several planned improvements on the horizon.

Future developments: taking advantage of the ongoing modernisation of the ENSDF format, additional delayed particle data – currently inaccessible via the IAEA API – will be incorporated, enhancing decay spectra completeness. To improve the accuracy and in-depth use of GENF/D, the uncertainty, and the error propagation, should be re-examined in the future. The first step would address the uncertainties on the Q-values, and its derived properties, in particular, to implement the mass covariance matrix provided by AME2020.

A major focus will be to revisit the heat computation, moving from the empirical Q/3 approximation for the beta decay average energy to more precise, isotope-dependent evaluations. The preliminary review of Section 4.1.3 suggests revision of the estimation of E_β from Q, with the use of a more refined empirical factor or by accounting for some dependence on the atomic number. Recently, new estimations of energies for light particle heat have been studied, especially with the proposal of advanced phenomenological beta model using Gamow-Teller decay strength [50]. A larger review of such approaches should yield more accurate decay heat predictions in future updates of GENF/D, where experimental observation is lacking.

Applications: GENF/D will become the backbone of TMX_Bateman, a Bateman equation solver, developed by Transmutex. This code is capable of computing the decay heat from heavy, light and electromagnetic decay heat as well over specific channels i.e., alpha, beta, gamma, electron heat. Special care is taken for gamma heat computation, and for the generation of the decay spectra. GENF/D will allow for better generation of clearance levels, raised by the radioprotection issues and during nuclear facility dismantling, and will also improve accuracy during fuel development.

Acknowledgments

The authors would like to thank K-H. Schmidt, for the help and discussions in the use of the GEF code, as well as X. Mougeot for the discussions, especially on BetaShape.

Funding

This work was supported by Transmutex, through the funding of a PhD, during a collaboration between Transmutex and CNRS.

Conflicts of interest

The authors declare that they have no competing interests to report.

Data availability statement

This article includes ENDF-6, as well as JSON and GNDS alternatives, of GENF/D.

Author contribution statement

Data Generation – Writing – Original Draft Preparation, Jonathan Collin, Abdel-Mjid Nourreddine; Writing – Review & Editing, Yves Schutz, Guilhem Lacaze, Donovan Maire.

References

All Tables

All Figures

	Fig. 1. Flowchart of data source for GENF/D generation.
In the text

	Fig. 2. Segrè diagram of available nuclides and their main decay modes in the GENF/D decay sub-library.
In the text

	Fig. 3. Log-scaled relative difference between ENDF/B original and recomputed from Q-value decay heat by atomic number Z. The discrepancies for the total decay heat (blue hexagon), as well as light particle (yellow diamond), electromagnetic (green square) heavy particle (grey bullet) decay heat are represented, with their global average.
In the text

	Fig. 4. Relative difference between GENF/D half-lives and JENDL-5.
In the text

	Fig. 5. Relative discrepancy between GENF/D and JEFF-4.0 for available decay energy (left) and neutron multiplicity (right).
In the text

Fig. 6. Log-scaled relative difference between ENDF/B decay heat estimates (considering 1023 W total heat of each radionuclides) and the one recomputed based on Table 6 convention, considering only the nuclides with spectral information. The discrepancies for the total decay heat (blue hexagon), as well as light particle (yellow diamond), electromagnetic (green square) heavy particle (grey bullet) decay heats are represented, with their global average.

In the text

	Fig. 7. Ratio of the average beta energy (from BetaShape) to the NUBASE2020 Q-value. The top panel shows results for β+ emissions, and the bottom panel for β− emissions. Data are shown for distinct isomeric states (I = 1 corresponds to first isomeric states, in darker tones, and I = 0 to ground states, in lighter tones), alongside the global average and a linear fit over atomic number Z.
In the text

	Fig. 8. Decay heat benchmark using ENDF/B-VIII.0 neutron cross-section for GENF/D against ENDF/B, JEFF and JENDL decay data (using their associated fission fragments). On the left, a long thermal irradiation on 235U. On the right, a thermal pulse experiment on 239Pu.
In the text

	Fig. 9. Decay heat benchmark using TENDL-2017 neutron cross-section, GEFY fission fragments and GENF/D against ENDF/B, JEFF and JENDL decay data. Five minutes irradiation at FNS of natNi, natMn, natFe, natCr.
In the text