Searcharxiv⌕ Search

arXiv · 2610.36154

RPAx Instabilities from Ground- and Excited-State References

Abstract

Random-phase approximation with exchange (RPAx) calculations based on ground- or excited-state references can yield complex excitation energies, causing the corresponding RPAx correlation energy to become complex as well. In this work, we investigate the origin of these complex energies using \ce{H2} in the 6-31G basis set as a minimal model. For the ground-state restricted Hartree-Fock reference, complex roots emerge at the well-known triplet instability associated with the appearance of a lower-lying unrestricted, symmetry-broken solution. For excited-state references, we identify two distinct mechanisms leading to complex excitation energies. The first is similarly associated with symmetry breaking. The second arises when a positive-energy excitation becomes degenerate with the positive-energy de-excitation associated with a negative-energy transition. Finally, we demonstrate the practical implications of these findings for spin-flip Bethe-Salpeter equation calculations of cyclobutadiene and state-specific core-ionization calculations for selected systems. These examples show that, in the absence of the relevant instabilities, methods sharing the generalized eigenvalue structure of RPAx can be applied to excited-state references without resorting to the Tamm-Dancoff or core-valence separation approximations, while retaining real excitation energies.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Loris Burth, Marios-Petros Kitsaras, Pierre-François Loos. 2026-10-26. RPAx Instabilities from Ground- and Excited-State References. https://arxiv.org/abs/2610.36154

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

KEEP EXPLORING

Related papers

Integer electron-count cuts in variational second-order reduced-density-matrix theory. I. Molecular dissociation with fixed real-space and orbital partitions

Van Aggelen et al. [Phys. Chem. Chem. Phys. 11, 5558 (2009)] showed that variational second-order reduced-density-matrix (2-RDM) calculations with the standard two-index conditions dissociate diatomic molecules into fractionally charged atoms with energies far below the exact limit, a drastic violation of size consistency; the proposed subsystem constraints require auxiliary calculations with the physical fragment Hamiltonians. We show that two long-known ingredients, combined, remove most of this failure without such calculations: the requirement that the number of electrons in a part of the molecule be an integer, imposed as linear conditions on the 2-RDM, and a chemically motivated partition of the molecule into atomic regions bounded at the bond critical point of the Hartree-Fock density. In a minimal basis, compared with full configuration interaction, a single such condition reduces the energy errors of NO$^+$ and CN$^-$ at 6 Å from 108.9 and 80.7 m$E_\mathrm{h}$ to less than 0.02 m$E_\mathrm{h}$. In N$_2^+$, whose fragments have the correct mean population but too small a fluctuation, the region and three-index conditions act complementarily, reducing the error at 4 Å from 138.0 to 1.6 m$E_\mathrm{h}$. At the separated-atom limit, the region cuts make the $P$, $Q$, $G$ energy additive over the fragments to within 0.2 m$E_\mathrm{h}$ in 7 of 10 systems, 3 already nearly additive without cuts; adding the $T_1$ and $T_2$ conditions also makes CO nearly additive, while a nonadditive error remains for N$_2^+$ and CO$^+$. Effective partitions are those whose condition is nearly saturated by the exact state.

physics.chem-ph↗

Full-data accuracy with fewer labels for training and fine-tuning machine-learning force fields

Machine-learning force fields (MLFFs) are reliable only near their training distribution, making efficient construction of diverse training sets a major bottleneck for both train-from-scratch and foundation fine-tuning workflows. Active learning can reduce this cost, but standard model-committee uncertainty is impractical for foundation MLFFs because each committee member requires a separate fine-tuning run. We present an active-learning workflow based on last-layer-projection regression (LLPR), a forward-pass-cheap per-configuration uncertainty estimator. Across molecular, condensed-phase, and electrolyte systems, LLPR identifies compact, high-value training sets that recover full-data accuracy using only a small fraction of electronic-structure labels. In foundation-model fine-tuning, LLPR-selected configurations approach the practical fine-tuning ceiling with substantially fewer labels than random selection. In iterative electrolyte fine-tuning, LLPR detects unphysical local coordination before new labelling, calibrates uncertainty in force-error units, and supports retrospective convergence monitoring. The resulting models reproduce reference density and Li--O coordination structure, providing a scalable uncertainty-quantification strategy across MLFF training regimes.

physics.chem-ph↗

Rotational-state-controlled dissociative ionization dynamics in $\mathrm{CF_2I_2}$

We investigated strong-field dissociative ionization of $\mathrm{CF_2I_2}$ ensembles prepared in different initial rotational-state distributions using an electrostatic deflector. Pronounced changes in ion-channel branching ratios revealed a strong dependence of the fragmentation dynamics on the initial rotational excitation. Analysis of fragment yields and their laser-power dependences identifies resonance-enhanced multiphoton ionization through an intermediate excited state, $\mathrm{CF_2I_2^{\ast}}$. The measured branching behavior indicates competition between stabilization into bound ionic states and dissociative channels, driven by near-threshold non-adiabatic Coriolis-type coupling. Tuning the rotational energy by only a few $μ$eV is sufficient to significantly alter the ionization dynamics and to redistribute the reaction products. These findings demonstrate the key role of rotational excitation in controlling non-adiabatic dynamics following strong-field ionization of $\mathrm{CF_2I_2}$.

physics.chem-ph↗