Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.23472

Single-electron fluorine bonding. II. Benchmarking Minnesota density functional approximations and popular ab initio methods for binding energies (BEs) and F-bond distances

Govinda Prasad Khanal·Elangannan Arunan
View original source ↗Read source document ↗Export metadata ↓

Abstract

Companion paper I (Khanal and Arunan, in this issue; arXiv:2610.16629 [physics.chem-ph]) reported reference BEs and F-bond distances for two new databases SEFB-15 (DF$\cdots$CH$_3$) and CSFB-14 (DF$\cdots$NH$_3$). Here, we assess performance of popular ab initio methods (MP2, MP3, MP4(DQ), MP4(SDQ), CCSD, and QCISD) with valence- and full-electron correlation treatments, and sixteen Minnesota DFAs, including M06-2X already assessed in companion paper I. CCSD/AVDZ reproduces reference F-bond distances closely, although BEs predicted by it deviate more. Among the 16 Minnesota DFAs, M08-HX reliably reproduces both reference values for DF$\cdots$CH$_3$ and DF$\cdots$NH$_3$, comparable to M05-2X and M06-2X for DF$\cdots$NH$_3$. Selected meta-GGAs (M06-L, M11-L, and MN15-L) exhibit spurious fractional charge transfer in cationic complexes even at infinite separation, whereas hybrid Minnesota DFAs do not, except M05 and M06 for NH$_3$F$^+\cdots$CH$_3$ and NH$_2$F$_2^+\cdots$CH$_3$. Although M08-HX shows slightly larger monomer delocalization error than M05-2X and M06-2X, it is more reliable for predicting BEs of SEFB complexes, indicating that monomer delocalization error alone does not govern DFA performance. Bader's intermolecular delocalization indices (DIs) further support M08-HX as more reliable than M05-2X and M06-2X for SEFB, while the larger DIs from M05-2X and M06-2X explain the degraded performance of their dispersion-corrected variants for DF$\cdots$CH$_3$, with only a minor impact for DF$\cdots$NH$_3$, highlighting the system-dependent nature of functional delocalization error. Present work reveals that M08-HX is more reliable than M05-2X and M06-2X for open-shell DF$\cdots$CH$_3$ complexes, while all three functionals show comparable reliability for closed-shell DF$\cdots$NH$_3$ complexes.

Explore related subjects

physics.chem-ph

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Govinda Prasad Khanal, Elangannan Arunan. 2026-10-16. Single-electron fluorine bonding. II. Benchmarking Minnesota density functional approximations and popular ab initio methods for binding energies (BEs) and F-bond distances. https://arxiv.org/abs/2610.23472

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

AceFF: A State-of-the-Art Machine Learning Potential for Small Molecules

We introduce AceFF, a pre-trained machine learning interatomic potential (MLIP) optimized for small molecule drug discovery. While MLIPs have emerged as efficient alternatives to Density Functional Theory (DFT), generalizability across diverse chemical spaces remains difficult. AceFF addresses this by refining the TensorNet2 architecture and training it on a comprehensive dataset of drug-like compounds. This approach yields a force field that balances high-throughput inference speed with DFT-level accuracy. We introduce two models sharing the same architecture: \mbox{AceFF-2}, trained on our own dataset of small molecules and supporting the essential medicinal chemistry elements (H, B, C, N, O, F, Si, P, S, Cl, Br, I); and \mbox{AceFF-2-OMOL}, trained on a subset of OMol25 containing the AceFF-2 elements plus the common biomolecular ion elements Na and K. The models are explicitly trained to handle charged states. Validation against rigorous benchmarks, including complex torsional energy scans, molecular dynamics trajectories, batched minimizations, and tests of force and energy accuracy, demonstrates that AceFF is state-of-the-art for organic molecules in the accuracy and speed regime important for drug discovery. The AceFF-2 and AceFF-2-OMol model weights and inference code are available at https://huggingface.co/Acellera/AceFF-2.0 and https://huggingface.co/Acellera/AceFF-2-OMOL.

physics.chem-ph↗
arXiv2026

Gap edge eigenpairs from density matrix purification using moments of the Dirac distribution

In this work, we propose a simple method to resolve the eigenstates located at the band gap edges of an electronic eigenspectrum using only the quasi-purified one-particle density matrix as input. The theoretical framework relies on the decomposition of the occupation number variance into a particle and a hole moment. These moments, when purified using power narrowing iterations, allow us to isolate the highest occupied and lowest unoccupied single-state projectors, \colp{giving ready access} to the corresponding eigenpairs. We demonstrate that when degeneracy is encountered, power narrowing remains able to deliver relevant mixed states. From a benchmark of selected molecules, we show that the method is robust and efficient since it requires from four to sixteen matrix-matrix multiplications in the worst case. The possibility of reducing the computational cost using the Lanczos subspace approach is discussed. The algorithm simplicity of power narrowing makes it very easy to implement in electronic structure codes or libraries already featuring Fermi operator expansion or density matrix purification.

physics.chem-ph↗
arXiv2026

Tersoff-NN: Hybrid Machine-Learned Bond-Order Extension of the Tersoff Potential

Accurate, robust, and efficient interatomic potentials are highly desirable for large-scale molecular dynamics simulations. Although machine-learning interatomic potentials can reproduce first-principles potential-energy surfaces with high accuracy, they typically incur higher computational costs than conventional empirical potentials and may be less reliable outside their training domains. We developed Tersoff-NN, a hybrid machine-learning potential for Si, which retains the analytical radial interactions and cutoff structure of the Tersoff potential while representing its environment-dependent bond-order factor using compact neural networks (NNs). Benchmarks for 32,768-atom systems show that Tersoff-NN is 4-12 times faster than the Behler-Parrinello-type NN potentials tested in this study. Tersoff-NN improves the short-range structure of liquid Si relative to the conventional Tersoff potential, closely reproduces the ab initio radial and angular distributions, and reproduces the equilibrium structure of diamond Si and qualitative energetic trends among several crystalline phases. Furthermore, melt-quench simulations reveal a disordered solid with predominantly four-coordinated, tetrahedral-like local environments. Additionally, characteristic reconstruction motifs are retained on Si surfaces despite their exclusion from the training dataset. A direct comparison with an alternative atom-centered bond-order NN highlights the critical role of physical constraints in ensuring dynamic stability. Despite exhibiting comparable validation errors, the selected triplet-based architecture remains stable, while an alternative model exhibits unphysical bond-order growth and atomic collapse. These results confirm that learning only the bond-order term improves the description of diverse Si environments while retaining substantial computational efficiency and enhancing dynamical stability.

physics.chem-ph↗