Electron Affinity of Actinide ( <scp>IV</scp> ) Carboxylate Complexes From <scp>MN12</scp> ‐L Density‐Functional Calculations and Explainable Machine Learning

ABSTRACT We present a relativistic density‐functional and explainable machine learning study of the vertical electron affinity (EA) of 14 homologous actinide(IV) carboxylate complexes [M(L) 3 ] + (M = Th–Lr; L = propionate, acrylate). All values were obtained at the MN12‐L level via the def2‐TZVPPD basis set for H, C, and O and a Stuttgart small‐core relativistic effective‐core potential, together with its associated valence basis, for the actinide. For every (metal, ligand) pair, all physically accessible An(IV) multiplicities were considered (52 calculations in total), and the ground state was assigned to the multiplicity with the lowest total electronic energy. The EA, evaluated as the negative of the lower α‐ and β‐LUMO eigenvalues (the “true LUMO”), increases monotonically across the series from 5.53 eV at Th(5f 0 ) to 10.80 eV at Lr(5f 13 ) for the propionates (5.95 → 10.58 eV for the acrylates), with one outlier: at Th, the LUMO is forced into the 7 s/7p–ligand manifold because no 5f acceptor is available. Squared coefficient population analysis of the corrected ground‐state wavefunctions confirms that for Pa–No, the LUMO is predominantly metal‐5f (59%–82% on the metal), whereas at Lr, the highest‐energy 5f spin orbital mixes strongly with the carboxylate π* manifold (5f ≈20%–28%, O ≈42%–50%). A nine‐feature explainable machine‐learning model trained on the corrected dataset reproduces the EA series under leave–one‐metal‐out cross‐validation with R 2 = +0.70 and MAE = 0.56 eV; permutation and SHapley additive exPlanations (SHAP) analyses identify the 5f‐electron count and the atomic spin–orbit constant as the single dominant predictor cluster (Pearson r = +0.94 for the 5f count). The cross‐ligand transferability is excellent ( R 2 = 0.85–0.88, MAE = 0.35–0.43 eV). The combined DFT–ML analysis provides a quantitative reference for An (IV) carboxylate EAs and isolates the 5f‐shell occupation as the single physical lever governing their reduction propensity.

Authors

Institutions

Publication Details

Journal
Journal of Computational Chemistry
Published
2026-07-17
DOI
https://doi.org/10.1002/jcc.70469
Primary Topic
Machine Learning in Materials Science
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Electron Affinity of Actinide ( IV ) Carboxylate Complexes From MN12 ‐L Density‐Functional Calculations and Explainable Machine Learning

Samy F. Mahmoud, Mohammad Abd Al-Hakim Badawi, Ali A. Khairbek, Renjith Thomas et al.
Journal of Computational Chemistry
Machine Learning in Materials Science
article

Electron Affinity of Actinide ( IV ) Carboxylate Complexes From MN12 ‐L Density‐Functional Calculations and Explainable Machine Learning

Samy F. Mahmoud, Mohammad Abd Al-Hakim Badawi, Ali A. Khairbek, Renjith Thomas, Ralph Puchta, Dalia I. Saleh
article en

Abstract

ABSTRACT We present a relativistic density‐functional and explainable machine learning study of the vertical electron affinity (EA) of 14 homologous actinide(IV) carboxylate complexes [M(L) 3 ] + (M = Th–Lr; L = propionate, acrylate). All values were obtained at the MN12‐L level via the def2‐TZVPPD basis set for H, C, and O and a Stuttgart small‐core relativistic effective‐core potential, together with its associated valence basis, for the actinide. For every (metal, ligand) pair, all physically accessible An(IV) multiplicities were considered (52 calculations in total), and the ground state was assigned to the multiplicity with the lowest total electronic energy. The EA, evaluated as the negative of the lower α‐ and β‐LUMO eigenvalues (the “true LUMO”), increases monotonically across the series from 5.53 eV at Th(5f 0 ) to 10.80 eV at Lr(5f 13 ) for the propionates (5.95 → 10.58 eV for the acrylates), with one outlier: at Th, the LUMO is forced into the 7 s/7p–ligand manifold because no 5f acceptor is available. Squared coefficient population analysis of the corrected ground‐state wavefunctions confirms that for Pa–No, the LUMO is predominantly metal‐5f (59%–82% on the metal), whereas at Lr, the highest‐energy 5f spin orbital mixes strongly with the carboxylate π* manifold (5f ≈20%–28%, O ≈42%–50%). A nine‐feature explainable machine‐learning model trained on the corrected dataset reproduces the EA series under leave–one‐metal‐out cross‐validation with R 2 = +0.70 and MAE = 0.56 eV; permutation and SHapley additive exPlanations (SHAP) analyses identify the 5f‐electron count and the atomic spin–orbit constant as the single dominant predictor cluster (Pearson r = +0.94 for the 5f count). The cross‐ligand transferability is excellent ( R 2 = 0.85–0.88, MAE = 0.35–0.43 eV). The combined DFT–ML analysis provides a quantitative reference for An (IV) carboxylate EAs and isolates the 5f‐shell occupation as the single physical lever governing their reduction propensity.

Journal of Computational ChemistryVol. 47(20)
Chandigarh University (IN), Latakia University (SY), Taif University (SA), Friedrich-Alexander-Universität Erlangen-Nürnberg (DE), Sacred Heart College (IN)
Taif University
Openalex Percentile: Top 18%
Machine Learning in Materials Science
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.