betaFeedback

WorksMathematics

A Coulomb ground-state density without Kohn-Sham ensemble representation

Abstract

We construct a finite three-electron Coulomb molecule whose spin-summed ground-state density cannot be reproduced by any ground-state ensemble of noninteracting electrons in a single real, spin-independent local potential in L3/2(R3)+L∞(R3)L^{3/2}(\mathbf R^3)+L^\infty(\mathbf R^3). The molecule has two equal positive integer nuclear charges, specified by an exact finite nonnumerical formula.

Introduction

The density-functional variational viewpoint originates in the work of Hohenberg and Kohn [9]. Uniqueness of a density–potential mapping and existence of a representing potential are separate questions; see the review [15] for their precise distinctions. Kohn–Sham density functional theory seeks to reproduce the density of an interacting electronic ground state by noninteracting electrons in a common local potential [11]. The constrained-search formulations of Levy and Lieb distinguish the existence of states with a prescribed density from the existence of a potential for which those states minimize the energy [13, 14]. Allowing ensembles enlarges the latter class: several degenerate noninteracting ground states may be mixed to obtain a density that is unavailable from one determinant. Ullrich and Kohn make this distinction explicit in their treatment of singlet ground-state ensembles [20]. Thus a failure of a particular orbital ansatz does not settle the ensemble representability question.

We consider the usual three-dimensional potential space

V=L3/2(R3;R)+L∞(R3;R).\mathcal{V} = L^{3/2}(\mathbb{R}^{3};\mathbb{R}) + L^{\infty}(\mathbb{R}^{3};\mathbb{R}).

For v∈Vv \in\mathcal{V}, the noninteracting Hamiltonian

Hs(v)=∑i=13(−12Δi+v(ri))H_s(v) = \sum_{i=1}^{3} \left(-\frac{1}{2}\Delta_i + v(r_i)\right)

acts on the antisymmetric space ⋀3L2(R3;C2)\bigwedge^{3} L^{2}(\mathbb{R}^{3};\mathbb{C}^{2}), with the same scalar potential for both spin components. Its realization is determined by its semibounded closed quadratic form. A ground-state ensemble is a positive trace-class operator Γ\Gamma of trace one whose range is contained in the eigenspace at inf⁡σ(Hs(v))\inf\sigma(H_s(v)). We denote its spin-summed one-body density by nΓn_{\Gamma}; it has integral three.

Theorem 1.1. Set D=10200D = 10^{200} and ez=(0,0,1)e_z = (0,0,1). There is a finite positive integer ZZ, specified by (97), for which the Coulomb Hamiltonian

HC=∑i=13(−12Δi−Z∣ri−(D/Z)ez∣−Z∣ri+(D/Z)ez∣)+∑1≤i<j≤31∣ri−rj∣H_C = \sum_{i=1}^{3}\left(-\frac{1}{2}\Delta_i - \frac{Z}{\lvert r_i - (D/Z)e_z\rvert} - \frac{Z}{\lvert r_i + (D/Z)e_z\rvert}\right) + \sum_{1\leq i<j\leq3}\frac{1}{\lvert r_i-r_j\rvert}

has a normalizable absolute ground state Ψ\Psi in the sector Sz=1/2S_z = 1/2, normalized by ∥Ψ∥2=1\lVert\Psi\rVert_2 = 1, but there is no v∈Vv \in\mathcal{V} and no ground-state ensemble Γ\Gamma of Hs(v)H_s(v) such that

nΓ=nΨalmost everywhere on R3.n_{\Gamma} = n_{\Psi}\qquad\text{almost everywhere on }\mathbb{R}^{3}.

@ԥҳа

The two nuclei have positive integer charges and distinct finite positions. The theorem concerns a fixed finite system. The exact specification of ZZ uses only spectral data, densities, and an angle minimization for a two-centre auxiliary problem; it does not test representability by other potentials. It is a nonnumerical specification, and no effective numerical bound for the charge is asserted.

Theorem 1.1 shows that a density of a finite interacting Coulomb ground state can fail to have a noninteracting ground-state ensemble representation in V\mathcal{V}. No restriction is placed on the degeneracy of a putative representing ground space or on the complex spinful states in its ensemble. The result makes no assertion about larger classes of distributional potentials, spin-dependent potentials, or nonlocal one-particle operators.

The same density also obstructs a local-potential first variation of the ensemble Hartree–exchange–correlation functional. More precisely, Corollary 8.1 excludes a potential in V\mathcal{V} representing its one-sided variation along every segment in the common finite-energy density domain.

Relation to earlier work

Explicit prescribed densities can already fail noninteracting vv-representability in V\mathcal{V}. Wagner and collaborators discuss a positive two-electron density with finite kinetic energy whose interior cusp forces its inverse potential outside this class [21], Section IV.G, eq:36. Their example is specified directly, without a construction as a finite Coulomb ground-state density.

Trushin, Erhard, and Görling report numerical indications of failures of ground-state noninteracting representability from atomic density inversions [19]. Open-shell inversion and symmetry restrictions are further investigated in [4]. The result here is an analytic exclusion for every ground-state ensemble of a specified finite Coulomb system, in the potential class V\mathcal{V}.

Existence results for other settings must be distinguished from this question. The theorem of Chayes, Chayes, and Ruskai concerns quantum lattice systems with a fixed bounded kinetic-plus-interaction operator [2]. Corso proves a rigorous Kohn–Sham formulation for spinless fermions on a bounded one-dimensional interval with generalized Neumann boundary conditions, allowing distributional potentials [3]. On a one-dimensional torus, Sutter and collaborators prove ensemble representability of strictly positive H1H^1 densities of prescribed particle number for spinful electrons, again allowing distributional potentials [18]. These results do not supply a representability theorem for three-dimensional Coulomb densities in V\mathcal{V}.

The asymptotic mechanism behind our construction was investigated by Gori-Giorgi, Gál, and Baerends [8]. When the leading removal amplitude has a nodal plane, coupling to other ionic channels can give the density the same exponential decay rate on and off that plane, with an additional polynomial suppression on the plane. Their analysis identifies the role of the angular part of the Laplacian and the resulting difficulties for a regular Kohn–Sham potential. Gori-Giorgi and Baerends subsequently studied the effective equation for the square root of the density and exact harmonically confined examples [7]. That effective equation is distinct from the common local potential determined below by two orbitals and an ensemble kinetic certificate. Here we prove the necessary estimates for a finite Coulomb system, and supply two further ingredients: a certificate valid against all ground-state ensembles and an obstruction to the entire sum space V\mathcal{V}.

The mechanism and the proof

Scale length by 1/Z1/Z. The nuclear charges become one, the half-separation becomes DD, and the electron repulsion has coefficient λ=1/Z\lambda= 1/Z. The two lowest one-particle orbitals are a positive even orbital bb and an odd orbital ee, positive above the median plane. Their splitting can be made arbitrarily small by increasing DD. At zero coupling, the three-electron ground determinant occupies b↑b \uparrow, b↓b \downarrow, e↑e \uparrow; its simple continuation in Sz=1/2S_z = 1/2 is the interacting state used below. Let nn denote the spin-summed density of this scaled interacting state.

Write E3,E2E_3,E_2 for the three-electron and ionic ground energies, and put

k=2(E2−E3),A(r)=e−krrβ,β=2−2λk−1.k = \sqrt{2(E_2-E_3)}, \qquad A(r) = e^{-k r}r^\beta, \qquad\beta= \frac{2-2\lambda}{k}-1.

The direct removal amplitude has a simple odd angular zero. In the complementary ionic space, the first nonzero planar term has amplitude of order λA(r)r−4\lambda A(r)r^{-4}. The full channel calculation proves that this coefficient is nonzero and controls the remainder uniformly through the region where it competes with the fast core term. Figure 1 locates the plane and the translated windows used to compare these asymptotics with an orbital equation.

A meridional section of the scaled molecule showing the nuclei, median plane, and translated window

Figure 1. A meridional section of the scaled molecule, drawn schematically and not to scale. The nuclei lie at ±Dez\pm D e_z; the horizontal line is the median plane z=0z=0. A bounded window is translated to a distant point pωp\omega in that plane. After normalization by (A(p)/p)2(A(p)/p)^2, the limiting density in the window vanishes quadratically on the plane, although the actual density there has a smaller nonzero tail. This distinction drives the limiting-orbital and trace arguments.

Trigonometric parametrizations also underlie constrained kinetic-energy formulations for paired real orbitals [12], Section 3.3. For the Coulomb density nn, we minimize

∫R3n∣∇θ∣2,∫R3nsin⁡2θ=1,\int_{\mathbb{R}^3} n\lvert\nabla\theta\rvert^2, \qquad \int_{\mathbb{R}^3} n\sin^2\theta= 1,

over odd angles, folded into [0,π/2][0,\pi/2] on the upper half-space. The orbitals

g=ncos⁡θ,h=nsin⁡θg = \sqrt{n}\cos\theta, \qquad h = \sqrt{n}\sin\theta

have squared norms two and one, respectively. They obey common formal-potential equations with energies zero and d>0d>0. A weighted spectral estimate proves that the paired/single determinant is a kinetic-energy minimum among all ensembles with density nn. Consequently, any representing local potential must agree, up to a constant, with V=Δg/(2g)V=\Delta g/(2g).

If V∈VV\in\mathcal{V}, translations of the orbital equations force every bounded limiting potential to equal d+k2/2d+k^2/2. This yields an exterior form bound and hence

earg∈H1(R3)for some a>k.e^{ar}g\in H^1(\mathbb{R}^3) \quad\text{for some } a>k.

The plane trace would then have the same exponential integrability. But θ=0\theta=0 on the plane, so g=ng=\sqrt{n} there, and the nonzero λA(r)r−4\lambda A(r)r^{-4} tail makes that trace integral diverge. This is the contradiction.

Several arguments have uses beyond this example. Keeping the unperturbed fast channel controls the crossover without tracking every perturbed excited threshold. Spin orthogonality prevents cancellation and makes the relevant relative constants independent of the small level splitting. The angle construction gives an ensemble kinetic certificate from a weighted Poincaré inequality, even before the formal potential is known to belong to a standard potential class. Finally, translated weak potential limits, local form compactness, and an H1H^1 trace convert a plane asymptotic into an obstruction to L3/2+L∞L^{3/2}+L^\infty, without assuming pointwise convergence of the potential at infinity. We give the estimates and their hypotheses explicitly.

Organization and conventions

Section 2 establishes the Coulomb spectral and perturbative facts. Section 3 proves the uniform density asymptotics and their fixed-parameter counterpart. Section 4 constructs the angle and proves the weighted gap. Section 5 establishes the full ensemble certificate and identifies a hypothetical representing potential. Section 6 proves the fixed-density obstruction (Theorem 6.4) and the qualitative Coulomb family (Theorem 6.5). Section 7 uses finite sufficient tests to select ZZ and complete Theorem 1.1. Section 8 derives the functional consequence for the physical density.

All inner products are linear in the second variable. Norms of ionic channel vectors are Hilbert-space norms. Spatial reflection means reflection in the median plane; a simultaneous reflection is applied to all electron coordinates when discussing a many-electron state. Limits uniform for small coupling always keep DD fixed first. Later fixed-parameter limits keep the selected positive coupling fixed.

The Coulomb system and its small-coupling limit

Let

h=L2(R3;C2),Hm=⋀mh\mathfrak{h}=L^2(\mathbb{R}^3;\mathbb{C}^2), \qquad\mathcal{H}_m=\bigwedge^m\mathfrak{h}

be the one-electron and antisymmetric mm-electron spaces. Spin is expressed in the basis ↑,↓\uparrow,\downarrow, and spatial reflections and rotations act trivially on spin. For D>0D>0, put

uD(x)=−1∣x−Dez∣−1∣x+Dez∣,hD=−12Δ+uD,ez=(0,0,1).(1)u_D(x)=-\frac{1}{\lvert x-D e_z\rvert}-\frac{1}{\lvert x+D e_z\rvert}, \qquad h_D=-\frac{1}{2}\Delta+u_D,\qquad e_z=(0,0,1). \tag*{(1)}

The corresponding mm-electron operators are

Hm,λ=∑i=1mhD(xi)+λ∑1≤i<j≤m1∣xi−xj∣,m=2,3,λ≥0.(2)H_{m,\lambda}=\sum_{i=1}^{m}h_D(x_i)+\lambda\sum_{1\leq i<j\leq m}\frac{1}{\lvert x_i-x_j\rvert}, \qquad m=2,3,\quad\lambda\geq0. \tag*{(2)}

They act in Hm\mathcal{H}_m. Hardy’s inequality in a particle coordinate, or in a relative coordinate, gives

∥∣xi−R∣−1ψ∥2+∥∣xi−xj∣−1ψ∥2≤C∥∇ψ∥2≤ε∥Δψ∥2+Cε∥ψ∥2.\bigl\|\lvert x_i-R\rvert^{-1}\psi\bigr\|_2+\bigl\|\lvert x_i-x_j\rvert^{-1}\psi\bigr\|_2\leq C\|\nabla\psi\|_2\leq\varepsilon\|\Delta\psi\|_2+C_\varepsilon\|\psi\|_2.

Thus these operators are self-adjoint on the antisymmetric part of H2(R3m;C2m)H^2(\mathbb{R}^{3m};\mathbb{C}^{2m}), and their form domain is the corresponding H1H^1 space. The same estimates show that λ↦Hm,λ\lambda\mapsto H_{m,\lambda} is an analytic family with this common operator domain. We write Em(λ)E_m(\lambda) for its ground energy whenever the ground level under consideration exists.

For a normalized three-electron state, its spin-summed density is

nΨ(x)=3∑σ1,σ2,σ3∫R6∣Ψ(x,σ1,y1,σ2,y2,σ3)∣2 dy1 dy2.(3)n_\Psi(x)=3\sum_{\sigma_1,\sigma_2,\sigma_3}\int_{\mathbb{R}^6}\lvert\Psi(x,\sigma_1,y_1,\sigma_2,y_2,\sigma_3)\rvert^2\,dy_1\,dy_2. \tag*{(3)}

In particular, ∫nΨ=3\int n_\Psi=3. Ensemble densities are defined by linearity, as detailed in Section 5.

Proposition 2.1 (Dilation to integer nuclear charges). Suppose that the density of a ground state of H3,1/ZH_{3,1/Z}, for a positive integer ZZ, has no noninteracting ground-state ensemble representation with a real spin-independent potential in L3/2(R3)+L∞(R3)L^{3/2}(\mathbb{R}^{3}) + L^{\infty}(\mathbb{R}^{3}). The physical Coulomb system with three electrons and charges Z,ZZ,Z at

(0,0,D/Z),(0,0,−D/Z)(0,0,D/Z), \qquad(0,0,-D/Z)

then has a ground-state density with the same obstruction.

Proof. Use physical coordinates ri\mathbf{r}_i and set xi=Zri\mathbf{x}_i=Z\mathbf{r}_i. The unitary dilation

(UZψ)(x1,…,xm)=Z−3m/2ψ(x1/Z,…,xm/Z)(U_Z\psi)(x_1,\ldots,x_m)=Z^{-3m/2}\psi(x_1/Z,\ldots,x_m/Z)

References

  1. [1]S. Agmon, Lectures on Exponential Decay of Solutions of Second-Order Elliptic Equations: Bounds on Eigenfunctions of N-Body Schrödinger Operators, Mathematical Notes, vol. 29, Princeton University Press, Princeton, NJ; University of Tokyo Press, Tokyo, 1982.
  2. [2]J. T. Chayes, L. Chayes, and M. B. Ruskai, Density functional approach to quantum lattice systems, Journal of Statistical Physics 38 (1985), 497–518. doi:10.1007/BF01010474.DOI
  3. [3]T. C. Corso, A rigorous formulation of density functional theory for spinless fermions in one dimension, Letters in Mathematical Physics 116 (2026), article 27. doi:10.1007/s11005-026-02051-1.DOI
  4. [4]J. Erhard, E. Trushin, and A. Görling, Kohn–Sham inversion for open-shell systems, The Journal of Chemical Physics 162 (2025), 034116. doi:10.1063/5.0239422.DOI
  5. [5]S. Fournais and T. Ø. Sørensen, Estimates on derivatives of Coulombic wave functions and their electron densities, Journal für die reine und angewandte Mathematik 775 (2021), 1–38. doi:10.1515/crelle-2020-0047. Preprint: arXiv:1803.03495v1 (2018).DOI
  6. [6]R. L. Frank and R. Seiringer, Non-linear ground state representations and sharp Hardy inequalities, Journal of Functional Analysis 255 (2008), 3407–3430. doi:10.1016/j.jfa.2008.05.015.DOI
  7. [7]P. Gori-Giorgi and E. J. Baerends, Asymptotic nodal planes in the electron density and the potential in the effective equation for the square root of the density, The European Physical Journal B 91 (2018), article 160. doi:10.1140/epjb/e2018-90225-3.
  8. [8]P. Gori-Giorgi, T. Gál, and E. J. Baerends, Asymptotic behaviour of the electron density and the Kohn–Sham potential in case of a Kohn–Sham HOMO nodal plane, Molecular Physics 114 (2016), 1086–1097. doi:10.1080/00268976.2015.1137643. Preprint: arXiv:1109.2204v2 (2015).
  9. [9]P. Hohenberg and W. Kohn, Inhomogeneous electron gas, Physical Review 136 (1964), B864–B871. doi:10.1103/PhysRev.136.B864.DOI
  10. [10]T. Kato, Perturbation Theory for Linear Operators, Classics in Mathematics, Springer-Verlag, Berlin–Heidelberg, 1995. Reprint of the corrected 1980 printing of the second edition. doi:10.1007/978-3-642-66282-9.DOI
  11. [11]W. Kohn and L. J. Sham, Self-consistent equations including exchange and correlation effects, Physical Review 140 (1965), A1133–A1138. doi:10.1103/PhysRev.140.A1133.DOI
  12. [12]D. Kumar and A. A. Ramabathiran, A direct approach to computing the non-interacting kinetic energy functional, preprint, 2025. arXiv:2507.00321v2.arxiv.org/abs/2507.00321
  13. [13]M. Levy, Universal variational functionals of electron densities, first-order density matrices, and natural spin-orbitals and solution of the v-representability problem, Proceedings of the National Academy of Sciences of the USA 76 (1979), 6062–6065. doi:10.1073/pnas.76.12.6062.DOI
  14. [14]E. H. Lieb, Density functionals for Coulomb systems, International Journal of Quantum Chemistry 24 (1983), 243–277. doi:10.1002/qua.560240302.DOI
  15. [15]M. Penz, E. I. Tellgren, M. A. Csirik, M. Ruggenthaler, and A. Laestadius, The structure of density-potential mapping. Part I: Standard density-functional theory, ACS Physical Chemistry Au 3 (2023), 334–347. doi:10.1021/acsphyschemau.2c00069.DOI
  16. [16]B. Simon, Semiclassical analysis of low lying eigenvalues. I. Non-degenerate minima: asymptotic expansions, Annales de l’Institut Henri Poincaré, Section A 38 (1983), 295–308. Journal archive.
  17. [17]L. Simon, Lectures on PDE, lecture notes, updated March 5, 2015. Author-hosted lecture notes.
  18. [18]S. M. Sutter, M. Penz, M. Ruggenthaler, R. van Leeuwen, and K. J. H. Giesbertz, Solution of the ν-representability problem on a one-dimensional torus, Journal of Physics A: Mathematical and Theoretical 57 (2024), 475202. doi:10.1088/1751-8121/ad8a2a.DOI
  19. [19]E. Trushin, J. Erhard, and A. Görling, Violations of the ν-representability condition underlying Kohn–Sham density-functional theory, Physical Review A 110 (2024), L020802. doi:10.1103/PhysRevA.110.L020802.DOI
  20. [20]C. A. Ullrich and W. Kohn, Kohn–Sham theory for ground-state ensembles, Physical Review Letters 87 (2001), 093001. doi:10.1103/PhysRevLett.87.093001.DOI
  21. [21]L. O. Wagner, T. E. Baker, E. M. Stoudenmire, K. Burke, and S. R. White, Kohn–Sham calculations with the exact functional, Physical Review B 90 (2014), 045109. doi:10.1103/PhysRevB.90.045109.DOI
  22. [22]Y. Yuan, Lecture 7: Calderón–Zygmund estimate, M583E, Linear and Nonlinear Elliptic Equations, Part I, University of Washington, May 22, 2019. Author-hosted lecture notes.

Paper details

vlnc:W_67frj4rp
Version
v1 · latest
Published
License
LicenseRef-OpenAI-public
Usage

views · 6

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

unreleased internal OpenAI model

View statement
Looking for related papers…
Contents