Research
Pair density Shannon information measures and their $N$-dependent behaviour in diatomic molecular systems
Overview Research area: Quantum information theory applied to quantum chemistry — specifically, Shannon entropy and mutual information of pair densities (two-electron distributions) in position and mo

- arXiv
- 2609.21260
- Published
- 2026-09-18
- Authors
- Saul J. C. Salazar, J. Antonio Zarate, J. M. Solano-Altamirano, Humberto G. Laguna, Julio M. Hernandez-Perez, Robin P. Sagar
AI summary
Overview
Research area: Quantum information theory applied to quantum chemistry — specifically, Shannon entropy and mutual information of pair densities (two-electron distributions) in position and momentum space for diatomic molecules.
Technical level: Intermediate. The paper assumes familiarity with Hartree-Fock theory, reduced/pair densities, and standard information-theoretic quantities, though its central questions are stated in accessible language.
Scope: The paper computes pair-density Shannon entropies and mutual informations for two diatomic series (homonuclear X–X and heteronuclear H–X) plus neutral atoms He–Ar at the Hartree-Fock level, tests whether the entropy sums grow with basis-set quality, checks whether position-space mutual information exceeds momentum-space mutual information, and fits three models to the N-dependence of the entropy and mutual-information sums.
What This Paper Is About
Information-theoretic measures such as Shannon entropy have been studied extensively for one-electron densities, but far less is known about the corresponding measures for pair densities, which encode electron correlation and chemical bonding. This paper calculates pair-density entropies and mutual information in both position and momentum space for a range of diatomic molecules, then asks how the resulting entropy sums and mutual-information sums behave as the number of electrons N grows, comparing three proposed functional forms.
Key Contributions
- Provides Hartree-Fock pair-density Shannon entropies (position S_Γ, momentum S_Π), entropy sums S_T, and mutual informations I_r and I_p for a homonuclear series (Li-Li through Ar-Ar, 16 molecules) and a heteronuclear series (H-H through H-Cl, 11 molecules), each in four basis sets (STO-3G, 3-21G, 6-31G, 6-31G(2df,p)).
- Tests the proposal that the pair entropy sum S_T acts as a wave-function-quality indicator, analogous to earlier claims for the one-electron entropy sum S_t.
- Systematically compares three N-dependent models — a + b ln N (from the Thomas-Fermi model), a ln I_1 + b ln N + c (a hydrogenic model using the first ionization potential), and a harmonic-oscillator-derived expression — against neutral atoms, homonuclear diatomics, and heteronuclear diatomics, using adjusted R² and the corrected Akaike Information Criterion (AICc).
- Extends the N-dependence analysis from one-electron entropy sums to pair entropy sums and mutual-information sums in molecular systems, which the authors state have not previously been explored for diatomics.
Main Findings
- Basis-set quality raises the pair entropy sum: S_T increases from STO-3G to the larger basis sets with one exception. The increase comes from S_Γ rising (delocalization in position space) while S_Π falls (localization in momentum space).
- Position-space mutual information always exceeds momentum-space: I_r > I_p for every diatomic system studied, matching prior results for atoms. The interpretation offered is that the exchange component of the pair density is relatively larger in position space, so the momentum pair density is closer to a separable Hartree-like reference.
- H₂ is the special case: H₂ has the smallest S_T of all systems and lies closest to the entropic uncertainty bound, since no reductions are needed to obtain its pair density. Its mutual information is exactly zero in both representations because the two electrons have opposite spins, so the exchange density vanishes on spin integration.
- Entropy sums rise with N, mutual information falls: In the X–X series S_T increases with N because more reductions from the N-particle density are required, adding uncertainty. Both I_r and I_p decrease with N, so the mutual-information sum I_T decreases with N in both series.
- All three models can fit the entropy sums: For S_t and S_T, all three functional forms represent the molecular data, with the middle (ionization-potential) model showing more structure in the plots because of its I_1 dependence.
- Only the oscillator-based model fits mutual information well: The first two models, which contain no explicit pair-density information, represent S_T reasonably but fail to capture I_T accurately; I_T fits have notably lower adjusted R² (e.g., 0.5581 for neutral atoms with a + b ln N).
- Parameters are transferable across chemical series: For S_t, the a + b ln N fit gives a = 6.2193, 5.9105, and 5.8795 for neutral atoms, X–X, and H–X respectively, all near the one-electron bound 3(1 + ln π) ≈ 6.43, with b near unity for atoms and increasing with chemistry (1.0775, 1.1431, 1.2046). For S_T, a = 12.2589, 11.7736, 11.6993 and b = 2.1974, 2.2967, 2.4212. This near-constancy across chemical environments is presented as evidence for a universal N-dependent behaviour.
- The third model gives the best adjusted R² but not the best AICc: The harmonic-oscillator model reaches adjusted R² = 0.9998 for S_t and S_T across all three series, yet AICc favours the ionization-potential model (e.g., −75.2179 versus −53.4676 for neutral-atom S_t). The α parameter is similar across species and varies inversely with N, while λ is positive, largest for neutral atoms, and shows the most variation — interpreted as capturing differing chemical environments.
- The oscillator formula is not an exact description: Because the fitted parameters shift noticeably between the k = 1 (S_t) and k = 2 (S_T) fits, the authors argue that real chemical systems require a more complex logarithmic argument than the oscillator expression provides.
- Not reported: The individual S_ρ and S_π values for the three series are stated to be available from the authors on request, and the third-model parameters for I_T are not present in the provided text.
Methodology in Plain English
The authors build Hartree-Fock wave functions for each molecule using the Gaussian 09 package, taking ground-state configurations and bond distances from standard spectroscopic compilations (Huber and Herzberg 1979; Hostutler et al. 2002). They then use a developer version of the DensToolKit code to evaluate the pair densities in position space and, via Fourier transform, in momentum space, and from these compute Shannon entropies, entropy sums, and mutual informations.
Mutual information is computed as the relative entropy between the true (non-separable) pair density and the product of its one-electron marginals — effectively a measure of how far the pair density departs from a Hartree-like, uncorrelated reference. Because Hartree-Fock pair densities contain only an exchange term beyond the separable part, non-zero mutual information in this framework traces back entirely to exchange and the Fermi hole.
Finally, they plot each information measure against the electron count N for three datasets — neutral atoms He–Ar, homonuclear diatomics, and heteronuclear diatomics — and fit three candidate formulas. Model quality is judged with the adjusted coefficient of determination and the corrected Akaike Information Criterion, the latter chosen because sample sizes are small and it penalizes extra parameters.
Why This Matters
Establishing what pair-density information measures actually quantify — delocalization, correlation, distance from a mean-field reference — gives chemists a rigorous vocabulary for describing bonding and exchange effects. If the N-dependence of entropy sums is truly universal, it provides a diagnostic baseline against which anomalous or strongly correlated systems could be spotted, and it connects molecular electronic structure to broader arguments about entropy bounds in fermionic and bosonic systems.
Real-world applications:
- Basis-set and wave-function quality assessment: S_T behaves as a quality indicator, offering a possible diagnostic alongside energy-based benchmarks in computational chemistry workflows.
- Pair-density functional theory development: Quantifying how much information the pair density carries beyond the one-electron density supports arguments for pair-based functionals.
- Correlation and exchange analysis in bonding: Mutual information gives a compact, basis-independent number summarizing how correlated electrons are in a given molecule, useful for comparing bonding environments.
- Continuous-variable quantum information and thermodynamics: The entropic uncertainty relations used here underpin work in continuous-variable quantum information, where position and momentum representations are the natural setting.
Industry relevance spans computational chemistry software vendors (basis-set and method validation), pharmaceutical and materials modelling groups that rely on Hartree-Fock and post-Hartree-Fock workflows, and quantum technology research that uses entropic uncertainty relations as figures of merit.
Future Directions
- Extending the analysis beyond Hartree-Fock to multi-determinant wave functions, which the paper flags as important in the ongoing discussion of indistinguishability and quantum entanglement.
- Testing and refining the N-dependent forms for pair densities: the authors show the oscillator expression is not exact for chemical systems, so a functional form with a more complex logarithmic argument — possibly incorporating one- and two-body potential effects — is needed.
- Applying the same protocol to a wider range of molecules (beyond diatomics) and to atoms beyond He–Ar, to test the claimed universality of the fitted parameters.
- Clarifying the physical meaning of the b parameter in a + b ln N, which the authors explicitly describe as an open question despite its near-constancy across series.
Target Audience
Quantum chemists and chemical physicists working on information-theoretic descriptions of electronic structure; researchers in density functional theory and pair-density functional theory; and quantum information scientists interested in continuous-variable uncertainty relations and correlation measures in realistic atomic and molecular systems. Graduate students with a background in Hartree-Fock theory will find the framing accessible, while the tabulated benchmark data will be of direct practical use to specialists.
Authors’ abstract
Pair density Shannon entropies and mutual information in position and in momentum space are calculated for two series of homonuclear and heteronuclear diatomic molecules using Hartree-Fock wave functions. The pair density entropy sum increases with the quality of the basis set. Mutual information, a measure of statistical correlation, is seen to be smaller in momentum space as compared to position space. The interpretation is that the momentum pair density is closer than the position pair density to a Hartree-like reference. The $N$-dependent behaviours of the entropy and mutual information sums are examined by fitting the data to three different model behaviours. Results show that these models are capable of representing the molecular data in differing degrees. The parameters obtained from a particular model are relatively constant across different chemical series. This adds evidence to the argument of a universal $N$-dependent behaviour of the entropy sums.