Simulating one hundred entangled atoms using projected-interacting full configuration interaction wavefunctions corrected by projected density functionals

Simulating entangled atoms is a prerequisite to modeling quantum materials and remains an outstanding challenge for theory. I introduce a correlated wavefunction approach capable of simulating large entangled systems, and demonstrate its application to a 300-electron active space. Projected-interacting full configuration interaction plus density functional theory PiFCI+DFT combines near-exact correlated wavefunctions of multiple partially-interacting model systems, each corrected by a formally exact density functional. This approach can access large active spaces and visualize entanglement and strong correlation while maintaining competitive accuracy for molecular properties.

💡 Research Summary

The manuscript introduces a novel correlated‑wavefunction methodology called Projected‑interacting Full Configuration Interaction plus Density Functional Theory (PiFCI+DFT) that enables the simulation of large, strongly‑entangled electronic systems with unprecedented efficiency and accuracy. The core idea is to decompose the full electron–electron interaction into a set of partially interacting reference systems, each defined by projecting the two‑electron operator onto a single many‑electron state |ϕₙ⟩. Because each projected reference requires at most two Slater determinants, the otherwise intractable Full‑CI problem for a 300‑electron active space collapses to a collection of 2 × 2 CI matrices.

Mathematically, the exact ground‑state energy can be expressed as a weighted sum of the energies of these reference systems, each corrected by a formally exact Hartree‑exchange‑correlation (HXC) functional of the same electron density. The Hohenberg–Kohn theorems guarantee the existence of such functionals, and the authors show that, if the exact functionals were available, any convex combination of the reference densities would reproduce the exact total energy and density. In practice, the authors adopt a “black‑box” approach: projection states are taken as the eigenvectors of the Kohn‑Sham density matrix projected onto atomic valence orbitals (STO‑3G basis), and the weights wₙ are chosen as the absolute overlaps |⟨ϕₙ|ϕₘ⟩|, ensuring normalization and avoiding double‑counting of correlation.

Two concrete functional choices are tested. PiFCI+HF treats all exchange exactly (Hartree‑Fock) and mimics a CASSCF‑like treatment of dynamical correlation. PiFCI+DFT employs a hybrid functional (25 % exact exchange) built from BLYP, LDA, D3 dispersion, and a projected‑LDA term for the projection states. Both schemes are combined with self‑consistent Kohn‑Sham orbitals and orbital energies, guaranteeing that all partially interacting references share the same uncorrelated electron density.

The method is benchmarked on several prototypical systems. For the nitrogen molecule (N₂), PiFCI+DFT reproduces the NEVPT2 reference curve: the equilibrium bond energy and geometry are essentially identical, and the dissociation limit correctly approaches zero, unlike standard DFT which over‑binds or under‑binds depending on the functional. A linear chain of ten nitrogen atoms (N₁₀) yields an equilibrium lattice spacing of 1.5 Å and a bond energy of ~25 kcal mol⁻¹, intermediate between isolated N₂ and a bulk lattice. The most striking result is the simulation of a 100‑atom nitrogen square lattice (N₁₀₀), representing a 300‑electron active space. PiFCI+DFT predicts an equilibrium spacing of 2.0 Å and a per‑atom bond energy of ~20 kcal mol⁻¹ (≈15 % of the N₂ bond), a chemically reasonable value that reflects the reduced coordination in the lattice. At dissociation, the entangled N₁₀₀ system’s energy matches that of 100 non‑interacting nitrogen atoms, confirming the method’s ability to capture both strong correlation and entanglement.

Computational cost is modest. For N₁₀₀, the algorithm diagonalizes 100 × 4×4 projected atomic density matrices, constructs 400 two‑electron CI blocks, and solves them. On a single 3.0 GHz AMD EPYC‑Milan core, a full PiFCI+HF energy evaluation completes in under 30 seconds, demonstrating scalability far beyond traditional multireference techniques, which often require hours to days for comparable systems.

Error statistics on subsets of the GMTKN55 benchmark show that PiFCI+DFT achieves a weighted mean absolute deviation comparable to dispersion‑corrected hybrid functionals (e.g., B3LYP‑D3) while simultaneously delivering low self‑interaction error and a “zero‑sum” behavior for correlation energy. Analysis of projected occupancies and full‑CI correlation energies distinguishes “normal” (e.g., H atom), “entangled” (e.g., H₂⁺), and “strongly correlated” (e.g., N₁₀₀) regimes, providing a quantitative diagnostic of electronic structure.

The authors discuss connections to existing approaches: CAS‑in‑DFT, MCPDFT, DMRG‑CAS, selected CI, local CI, DFT+U, and symmetry‑breaking/restoration methods. PiFCI+DFT can be viewed as a unifying framework that incorporates the strengths of multireference wave‑function theory (exact treatment of selected interactions) and density‑functional theory (efficient handling of the bulk electron density). It can be combined with DMRG for larger active spaces, with local CI for spatially localized interactions, or with DFT+U‑type corrections for Hubbard‑like models, offering a flexible platform for future methodological development.

In summary, PiFCI+DFT provides a formally exact, black‑box, and computationally affordable route to simulate systems with hundreds of entangled electrons. By projecting electron–electron interactions onto a manageable set of reference states and correcting each with an appropriate density functional, the method captures both static (multireference) and dynamic correlation, reproduces dissociation limits, and yields chemically sensible energetics for extended lattices. This breakthrough opens the door to realistic quantum‑material simulations, high‑temperature superconductors, quantum sensors, and other applications where large‑scale electronic entanglement plays a pivotal role.