Why Molecular Schrödinger Equations Are Unsolvable Exactly

Every quantitative prediction in modern computational chemistry — a drug candidate's binding energy, a catalyst's activation barrier, a battery electrolyte's redox potential — is computed using a method built specifically to manage the non-separability demonstrated here. Density functional theory (DFT), used to screen candidate drug molecules and battery materials, replaces the intractable $\sum_{i<j}1/r_{ij}$ with a functional of the electron density that is far cheaper to evaluate while trying to capture the same physics; its approximations (the "exchange-correlation functional") are, in essence, engineered stand-ins for the exact but unreachable $1/r_{12}$ physics. Coupled-cluster theory, the gold standard for small-molecule thermochemistry used in atmospheric chemistry and combustion modeling, is a systematic hierarchy for recovering correlation energy order by order, converging (in principle, and often in practice) toward the exact answer that direct solution can never provide. In materials science, quantum Monte Carlo methods sample the many-electron wavefunction stochastically precisely because no closed-form or even efficient deterministic representation of the exact correlated wavefunction exists for solids. Even the entire practice of building molecular orbital diagrams in introductory organic chemistry — assigning electrons to individual, one-electron orbitals — is itself the mean-field simplification born from this same unsolvability, inherited all the way down into the language chemists use every day to talk about bonding.

Ask the tutor: How does the Hartree-Fock method turn the non-separable electron repulsion term into a separable effective potential, step by step?

Consider first why the hydrogen atom is solvable at all. Its Hamiltonian, after separating the center-of-mass motion, is

a single kinetic energy operator plus a single potential energy operator that depends on the coordinates of exactly one particle (the electron relative to the nucleus). Because the potential has spherical symmetry, this equation separates in spherical coordinates into radial and angular factors, giving the exact hydrogenic orbitals $\psi_{nlm}(r,\theta,\phi)$ and the exact energies $E_n = -13.6\,\text{eV}/n^2$. This is the only class of atomic problem where an exact closed-form solution exists, and it is solvable precisely because there is only one particle whose position matters to the potential energy. Now go to helium: two electrons (labeled 1 and 2) orbiting a nucleus of charge $Z=2$. The full non-relativistic, fixed-nucleus Hamiltonian is

The first four terms are exactly the kind of one-particle operators that appear in hydrogen: each depends only on $\mathbf{r}_1$ or only on $\mathbf{r}_2$. If the Hamiltonian stopped there, the Schrödinger equation would separate immediately, and helium would have an exact, hydrogen-like solution. It is the fifth term, $e^2/(4\pi\varepsilon_0 r_{12})$ — with $r_{12}=|\mathbf{r}_1-\mathbf{r}_2|$ — that ruins everything. This term depends on both electrons' positions simultaneously and cannot be decomposed into a piece belonging to electron 1 plus a piece belonging to electron 2. It is the mathematical signature of the physical fact that where electron 1 is changes the potential felt by electron 2, and vice versa: their motions are correlated.

The physical origin of the trouble is Coulomb's law itself: the repulsion energy between two charges depends on their separation, not on either charge's position in isolation. Separation of variables requires the potential to be additively decomposable, $V(\mathbf{r}_1,\mathbf{r}_2) = f(\mathbf{r}_1)+g(\mathbf{r}_2)$; the function $1/|\mathbf{r}_1-\mathbf{r}_2|$ is not of this form for any choice of $f$ and $g$, and no rotation, translation, or clever choice of coordinates changes that fact, because the non-separability is intrinsic to the functional form, not an artifact of the coordinate system. It is worth being precise about what does not cause the problem, because these are common false suspects. It is not the number of electrons per se — a hypothetical "helium" with two non-interacting electrons (no $1/r_{12}$ term) would separate perfectly into a product of two hydrogen-like $Z=2$ orbitals, with energy exactly $E = -Z^2(13.6\,\text{eV})\times(1/1^2+1/1^2) = -108.8\,\text{eV}$. It is not electron spin, which enters only through antisymmetrization (the Pauli Exclusion Principle) and does not by itself prevent separability. It is not nuclear motion — that is a separate, solvable complication handled by The Born-Oppenheimer Approximation. The single, irreducible culprit is the electron-electron Coulomb repulsion operator.

Every additional electron adds not one but several new coupling terms: an $n$-electron system has $\binom{n}{2} = n(n-1)/2$ distinct electron pairs, each contributing its own $1/r_{ij}$ term. Lithium already has three such terms ($r_{12}, r_{13}, r_{23}$); a modest ten-electron molecule like methane has 45. The non-separability does not merely persist as systems grow — it compounds combinatorially.

The general $N$-electron, fixed-nuclei ("clamped nucleus" or Born-Oppenheimer electronic) Hamiltonian in atomic units is

where:

• $-\tfrac12\nabla_i^2$ is the kinetic energy operator for electron $i$,
• $Z_A$ is the charge of nucleus $A$ and $r_{iA}$ the distance from electron $i$ to nucleus $A$ (one-electron operators, sum decomposes cleanly),
• $r_{ij}=|\mathbf r_i-\mathbf r_j|$ is the distance between electrons $i$ and $j$, and the final double sum runs over all unique electron pairs $j>i$.

Reading: the first two sums are collections of independent one-electron operators — each acts on only one electron's coordinates, exactly as in the hydrogen atom, so together they would separate. The third sum, the electron repulsion term $\hat{V}_{ee}$, is a genuinely two-electron (in general many-body) operator: it cannot be rewritten as a sum of one-electron pieces. This is the operator responsible for everything that follows in quantum chemistry. The separability criterion can be stated as a clean mathematical test. A Hamiltonian $\hat H(\mathbf r_1,\ldots,\mathbf r_N)$ admits a product (separable) solution $\Psi = \prod_i \psi_i(\mathbf r_i)$ with $\hat H\Psi = E\Psi$ reducing to $N$ independent one-particle equations if and only if $\hat H$ can be written as

where:

• $\hat h_i$ is a one-electron operator depending only on the coordinates (and momenta) of electron $i$.

Reading: this is precisely the structure of a sum of independent one-dimensional harmonic oscillators, or of the free-particle-in-a-box in 3D, or of hydrogen's kinetic-plus-Coulomb operator. Whenever $\hat H$ has this additive structure, substituting the product ansatz into $\hat H\Psi=E\Psi$ and dividing through by $\Psi$ turns the eigenvalue problem into $N$ decoupled single-variable equations, each solvable independently, with $E = \sum_i \varepsilon_i$. The many-electron Hamiltonian fails this test because of $\hat V_{ee}=\sum_{i<j}1/r_{ij}$, a genuinely two-electron operator: no matter how the coordinates are relabeled or transformed, $1/r_{ij}$ cannot be written as $u(\mathbf r_i)+w(\mathbf r_j)$, because it is a function of the difference $\mathbf r_i - \mathbf r_j$ that does not decompose additively (this is proven rigorously below).

where:

• $\hat h_i = -\tfrac12\nabla_i^2 - \sum_A Z_A/r_{iA}$ collects the kinetic energy and nuclear attraction for electron $i$ alone.

Reading: this decomposition is exactly the starting point of the Hartree-Fock method: it isolates the separable "core" of the Hamiltonian from the one troublesome piece, and every approximation strategy in quantum chemistry amounts to a different way of approximating $\hat V_{ee}$ by something that is separable (a mean field), then correcting for the difference.

Example 1 — Foundational: showing the non-interacting helium Hamiltonian separates exactly. Suppose (unphysically, for illustration) that the two electrons in helium did not repel each other — imagine switching off $1/r_{12}$. The Hamiltonian becomes

with $Z=2$. This is manifestly of the separable form $\sum_i \hat h_i$, so we substitute the product ansatz $\Psi_0 = \psi_{1s}(\mathbf r_1)\psi_{1s}(\mathbf r_2)$, where $\psi_{1s}$ is the exact hydrogenic $1s$ orbital scaled to $Z=2$:

Each hydrogenic energy level for nuclear charge $Z$ and principal quantum number $n=1$ is $\varepsilon_n = -Z^2(13.6\,\text{eV})/n^2$, so with $Z=2, n=1$:

This is an exact result for the fictitious non-interacting system — no approximation was used, because the Hamiltonian truly is separable once $1/r_{12}$ is deleted. The experimental ground-state energy of real helium is $-79.0\,\text{eV}$ (from ionization energies: first IE $= 24.6\,\text{eV}$, second IE $=54.4\,\text{eV}$, sum $=79.0\,\text{eV}$ below the doubly-ionized reference). The discrepancy, $-108.8-(-79.0)=-29.8\,\text{eV}$, is entirely the electron-electron repulsion energy that was switched off — a huge effect, roughly 27% of the total binding energy, underscoring that $1/r_{12}$ is not a minor perturbation but a first-order-important term that happens to also be exactly the term that prevents exact solution. Limiting-case check: as $Z\to\infty$ (a very highly charged nucleus), the nuclear attraction terms $-Z/r_i$ dominate over the fixed-strength repulsion $1/r_{12}\sim O(1)$, so the relative importance of the non-separable term shrinks — the system becomes better and better approximated by the separable, non-interacting picture. This matches physical intuition: highly charged nuclei pull electrons in so tightly and closely to the nucleus that electron-electron repulsion becomes a small correction, which is why hydrogenic (independent-particle) approximations work increasingly well for highly ionized heavy atoms (e.g., Fe$^{24+}$) even though they fail badly for neutral atoms.

Example 2 — Applied: estimating the correlation energy hidden in $1/r_{12}$ for the real helium atom. Quantum chemists define the correlation energy as the difference between the exact non-relativistic energy and the best possible energy obtainable from a single product-like (Hartree-Fock, mean-field) wavefunction — i.e., the energy contribution attributable specifically to the non-separable part of the Hamiltonian that mean-field theory cannot capture. For helium, high-precision non-relativistic calculations give the exact ground-state energy as

The best possible Hartree-Fock (single-product, mean-field) energy — which replaces $1/r_{12}$ with an averaged, separable effective potential and therefore is separable by construction — gives

The correlation energy is the gap between these:

This is small compared to the total binding energy (about 1.4%) but it is exactly the part of the physics that a separable (product) wavefunction cannot describe by definition — it exists solely because $\Psi(\mathbf r_1,\mathbf r_2)$ is not truly a product function, i.e., because electron 1's position is statistically correlated with electron 2's position due to their mutual repulsion (they tend to avoid each other more than an averaged mean-field description predicts). Though numerically modest for helium, correlation energy is chemically decisive: it can be comparable to or exceed reaction barrier heights and bond dissociation energy differences, which is why methods like configuration interaction, coupled-cluster theory, and Møller-Plesset perturbation theory exist purely to recover it. Limiting-case check: in the limit of a one-electron system (any hydrogenic ion), there are zero electron pairs, so $E_{\text{corr}}\equiv 0$ identically — Hartree-Fock is exact for one electron, consistent with the fact that the non-separable repulsion term simply does not exist when $N=1$.

Example 3 — Exam/competition level: quantifying non-separability via a perturbative first-order estimate, and comparing to the exact result. A standard first approximation treats $1/r_{12}$ as a perturbation on top of the separable zeroth-order Hamiltonian from Example 1, using first-order perturbation theory (a technique whose entire justification is that the unperturbed problem is exactly separable while the full problem is not). The first-order correction to the ground-state energy is the expectation value of the perturbation in the unperturbed (separable) ground state:

This integral (the "Coulomb integral" $J_{1s,1s}$) has a well-known closed form for hydrogenic $1s$ orbitals with effective charge $Z$:

For helium, $Z=2$:

So the first-order-corrected energy is

Compare to the exact experimental value $-79.0\,\text{eV}$: the first-order perturbative treatment, which explicitly builds in the non-separable term as a correction rather than ignoring it, closes most of the gap between the crude non-interacting estimate ($-108.8\,\text{eV}$) and reality, but overshoots slightly, landing $4.2\,\text{eV}$ above the true value. The remaining error comes from second- and higher-order perturbative corrections and from the fact that first-order perturbation theory still uses the unperturbed, separable orbitals rather than allowing them to relax and correlate in response to $1/r_{12}$ — a limitation directly traceable to the fact that $1/r_{12}$ genuinely mixes the coordinates and a single perturbative order cannot fully capture that mixing. Limiting-case check: setting $Z\to 0$ (an unphysical, unbound limit) sends $J_{1s,1s}\to 0$, correctly reflecting that with no nuclear attraction there is no bound $1s$ orbital to evaluate the repulsion integral over — the formula's dependence on $Z$ is dimensionally and physically sensible, scaling linearly with nuclear charge as expected since the orbital itself contracts as $Z$ increases ($\langle r\rangle \propto 1/Z$), making the average electron-electron distance shrink and the repulsion integral grow proportionally.

The person most associated with confronting this problem head-on, and turning "unsolvable" into "approximately solvable," is Douglas Rayner Hartree. Working in the late 1920s, motivated directly by the failure of exact methods beyond hydrogen, Hartree developed the self-consistent field method: replace the intractable instantaneous repulsion $1/r_{12}$ with an averaged potential that each electron feels due to the smeared-out charge distribution of all the others. This averaged potential depends only on one electron's coordinate at a time — it is separable by construction — turning the impossible many-body problem into a solvable, iterative sequence of one-electron problems whose solutions must be self-consistent with the averaged field they generate. Hartree's method (later corrected for antisymmetry by Vladimir Fock, giving Hartree-Fock theory) is the direct historical and logical descendant of the non-separability problem demonstrated in this article: it is quite literally an engineered approximation whose entire purpose is to make $\hat V_{ee}$ separable at the cost of losing correlation energy, exactly as quantified in Example 2.