QTAIM bond-critical-point topology beyond Cohen scaling in tetrahedral semiconductors: data, provenance and code (V9.3).

Reproducibility package for the manuscript "Bonding descriptor or stiffness predictor? A cross-validated test of QTAIM bond-critical-point topology beyond Cohen scaling in tetrahedral semiconductors" (H. Aourag, manuscript version V9.3).

The study tests whether two QTAIM descriptors of the bond critical point (BCP) carry information on the bulk modulus beyond Cohen-type bond-length scaling and its family-ionicity correction, B = (a + bλ) d^-3.5. The two descriptors are the density ρ_BCP and the displacement A_BCP = |s − 0.5|. The test uses out-of-sample cross-validation and residual-association tests on 32 tetrahedral semiconductors.

Contents:

• QTAIM descriptors for 36 tetrahedral semiconductors, computed on the PAW all-electron (valence + core) density: Quantum ESPRESSO 6.7MaX, PBE, pp.x plot_num = 21, Critic2, with a locked geometric BCP selection protocol.
• Full DFT/QTAIM provenance: Quantum ESPRESSO and Critic2 inputs and outputs for all production materials, and the PAW datasets used, identified by MD5 checksum.
• A provenance-locked recalculation of C, Si, SiC, BN, GaAs and ZnSe. The structures are taken from the same Materials Project ElasticityDoc records that supply the K_VRH values (ΔK = 0).
• A grid-convergence study of BN and AlN (ecutrho = 480–1920 Ry). The production grid (960 Ry) agrees with the finest grid within the pre-specified tolerances.
• A density-representation pilot comparing plot_num 0, 17 and 21 for AlN, GaN and BeS.
• Python scripts that regenerate Tables 1–4, Figures 1–4 and all 469 numerical values printed in the manuscript, with an automatic manuscript–data consistency check. Running python run_all.py ends with "ALL PASS".
• The pre-specified provenance-freeze rules and the changelog of every modified value.

Main result: within this cohort, neither ρ_BCP nor A_BCP improves the Cohen–λ model out of sample. None of the 24 residual-association tests survives Holm correction.

Licences: data, provenance files and documentation under CC BY 4.0; source code (*.py) under MIT.