Born–Oppenheimer PES

Almost every note in this garden quietly assumes one object: a single scalar function $E(\{\mathbf{R}\})$ of the nuclear coordinates, the potential energy surface (PES) on which nuclei move, relax, and vibrate. Phonons are its curvature, forces are its gradient, structural transitions are its topography. That function is not handed to us by the Schrödinger equation — it is the output of the Born–Oppenheimer (BO) adiabatic separation, the controlled decoupling of fast electrons from slow nuclei. This note states the full molecular Hamiltonian, makes the separation precise via the mass ratio, defines the PES as the eigenvalue of the clamped-nuclei electronic problem plus nuclear repulsion, and — because BO is an approximation — marks its failure regime (conical intersections, near-degeneracy, strong electron–phonon coupling) honestly.

Prerequisites

the many-electron Schrödinger equation · the variational principle (Rayleigh–Ritz)

1. The molecular Hamiltonian

Take $N_e$ electrons at positions $\mathbf{r}=\{{\mathbf{r}}_i\}$ and $N_n$ nuclei of charges $Z_A$ and masses $M_A$ at positions $\mathbf{R}=\{{\mathbf{R}}_A\}$. In the non-relativistic limit the full Coulomb Hamiltonian is

$$H=T_n+T_e+V_{ne}+V_{ee}+V_{nn},$$

with each term in operator form (atomic units $\hslash =m_e=e=4\pi {\epsilon }_0=1$, so that the electron mass is unity and nuclear masses $M_A$ are in units of $m_e$):

$$\begin{array}{rlrl}{T}_n& =-\sum _{A=1}^{N_n}\frac{1}{2M_A}{\nabla }_{{\mathbf{R}}_A}^2,& T_e& =-\sum _{i=1}^{N_e}\frac{1}{2}{\nabla }_{{\mathbf{r}}_i}^2,\\ V_{ne}& =-\sum _{i=1}^{N_e}\sum _{A=1}^{N_n}\frac{Z_A}{|{\mathbf{r}}_i-{\mathbf{R}}_A|},& V_{ee}& =\frac{1}{2}\sum _{i\ne j}\frac{1}{|{\mathbf{r}}_i-{\mathbf{r}}_j|},\\ V_{nn}& =\frac{1}{2}\sum _{A\ne B}\frac{Z_A{Z}_B}{|{\mathbf{R}}_A-{\mathbf{R}}_B|}.& & \end{array}$$

Every term is exact and every term couples electrons to nuclei except $T_e$, $V_{ee}$ (purely electronic) and $T_n$, $V_{nn}$ (purely nuclear). The single operator that entangles the two sets is the electron–nucleus attraction $V_{ne}$; it is what makes the electronic energy depend on the nuclear geometry, and the BO program is the statement that we may treat that dependence parametrically. The molecular eigenproblem $H\Psi (\mathbf{r},\mathbf{R})=\mathcal{E}\Psi (\mathbf{r},\mathbf{R})$ is a partial differential equation in $3(N_e+N_n)$ coordinates and is hopeless as written; the separation below is what makes it tractable.

2. Adiabatic separation and the mass ratio

The physical input is the disparity of scales: $m_e/M_A\ll 1$ for every nucleus (the lightest, hydrogen, is already $\sim 1836$). Electrons are light and fast; nuclei are heavy and slow. On the timescale over which a nucleus moves appreciably, the electronic cloud relaxes essentially instantaneously to the ground state of the current nuclear configuration — the electrons follow the nuclei adiabatically, never being left in an excited state by the nuclear motion. Equivalently, the off-diagonal couplings that would mix electronic states scale as a positive power of $m_e/M_A$ and can be ordered out.

Born and Oppenheimer made this rigorous as an expansion in the small parameter

$$\kappa ={\left(\frac{m_e}{M}\right)}^{1/4},$$

with $M$ a representative nuclear mass. The quartic root is the natural bookkeeping parameter: nuclear vibrational quanta scale as ${\kappa }^2$ (so ${\omega }_{\text{vib}}\sim {\kappa }^2\Delta E_{\text{elec}}$, the familiar $\sqrt{m_e/M}$ separation of electronic and vibrational energy scales), nuclear displacements about equilibrium as $\kappa$, and the leading non-adiabatic corrections enter at ${\kappa }^4=m_e/M$. The harmonic lattice — phonons — emerges precisely at order ${\kappa }^2$ in this expansion; the rigorous accounting is App. VIII of Born & Huang. The smallness of $\kappa$ is why the surface picture works, and the order at which it first fails (next section, $\S 6$) is set by the same parameter.

3. The electronic problem (clamped nuclei)

Freeze the nuclei at a configuration $\mathbf{R}$ — set $T_n=0$ — and define the electronic Hamiltonian as everything that remains and acts on the electronic coordinates:

$$H_{\text{elec}}(\mathbf{R})=T_e+V_{ne}(\mathbf{r};\mathbf{R})+V_{ee}=-\sum _i\frac{1}{2}{\nabla }_{{\mathbf{r}}_i}^2-\sum _{i,A}\frac{Z_A}{|{\mathbf{r}}_i-{\mathbf{R}}_A|}+\frac{1}{2}\sum _{i\ne j}\frac{1}{|{\mathbf{r}}_i-{\mathbf{r}}_j|}.$$

The nuclear positions $\mathbf{R}$ enter only as parameters (through $V_{ne}$), not as dynamical variables — this parametric dependence is the heart of the BO picture and the reason for the semicolon in the notation. For each fixed $\mathbf{R}$ this is a stationary many-electron Schrödinger equation,

$$H_{\text{elec}}(\mathbf{R}){\psi }_{\text{elec}}^{(n)}(\mathbf{r};\mathbf{R})=E_{\text{elec}}^{(n)}(\mathbf{R}){\psi }_{\text{elec}}^{(n)}(\mathbf{r};\mathbf{R}),$$

whose eigenfunctions ${\psi }_{\text{elec}}^{(n)}$ depend on the electron coordinates $\mathbf{r}$ and parametrically on $\mathbf{R}$, and whose eigenvalues $E_{\text{elec}}^{(n)}(\mathbf{R})$ are smooth functions of geometry (away from the degeneracies of $\S 6$). For each nuclear configuration one solves a different electronic problem; sweeping $\mathbf{R}$ traces out the surface. The eigenstates at fixed $\mathbf{R}$ are complete on the electronic Hilbert space, which is what licenses the adiabatic (Born–Huang) expansion of the exact molecular wavefunction, $\Psi (\mathbf{r},\mathbf{R})={\sum }_n{\chi }_n(\mathbf{R}){\psi }_{\text{elec}}^{(n)}(\mathbf{r};\mathbf{R})$, with nuclear amplitudes ${\chi }_n$; the BO approximation is the truncation of this sum to a single electronic surface.

4. The variational ground state (Rayleigh–Ritz)

In practice ${\psi }_{\text{elec}}^{(0)}$ is never obtained exactly. The working definition of the ground electronic state is variational: by the Rayleigh–Ritz principle the energy functional

$$E[\psi ]=\frac{⟨\psi |H_{\text{elec}}(\mathbf{R})|\psi ⟩}{⟨\psi |\psi ⟩}\ge E_{\text{elec}}^{(0)}(\mathbf{R})$$

is bounded below by the true ground-state eigenvalue for any trial $\psi$, with equality iff $\psi$ is the exact ground state. The ground state is the stationary point,

$$\delta E[\psi ]=0,$$

over the admissible (normalized, antisymmetric) trial space. This is the principle that every electronic-structure method instantiates — Hartree–Fock minimizes over single Slater determinants, configuration interaction over fixed-length CI expansions, and Kohn–Sham DFT recasts the search as a constrained minimization over densities (the variational object becomes the Kohn–Sham orbitals; the upper-bound guarantee is what the chosen functional space inherits). The surface a calculation actually delivers is the variational $E_{\text{elec}}(\mathbf{R})$ for the chosen ansatz — exact only in the complete-basis, exact-functional limit.

5. The potential energy surface

Add back the nuclear–nuclear repulsion, which is a pure function of $\mathbf{R}$ (a constant from the electrons’ point of view at fixed geometry). The Born–Oppenheimer potential energy surface for nuclear motion on the $n$-th electronic state is

$$\boxed{E^{(n)}(\mathbf{R})=E_{\text{elec}}^{(n)}(\mathbf{R})+V_{nn}(\mathbf{R})}$$

and the nuclei then move on this single scalar surface, governed by the nuclear Schrödinger equation $[T_n+E^{(n)}(\mathbf{R})]{\chi }_n(\mathbf{R})=\mathcal{E}{\chi }_n(\mathbf{R})$, in which $E^{(n)}(\mathbf{R})$ plays the role of the effective potential — the surface is the potential. This object is the root of essentially the whole lattice-dynamics and structure branch:

• Equilibrium geometry is a minimum of $E(\mathbf{R})$: $\partial E/\partial {\mathbf{R}}_A=0$. The gradient is the force on a nucleus, ${\mathbf{F}}_A=-\partial E/\partial {\mathbf{R}}_A$, and the Hellmann–Feynman theorem evaluates it as an expectation value over the electronic state without differentiating the wavefunction.
• Vibrations / phonons are the curvature of the surface at that minimum. The harmonic force constants are the Hessian ${\Phi }_{A\alpha ,B\beta }={\partial }^{2}E/\partial R_{A\alpha }\partial R_{B\beta }$, and the dynamical matrix is its mass-weighted Fourier transform, $D_{A\alpha ,B\beta }=(M_A{M}_B{)}^{-1/2}{\partial }^{2}E/\partial R_{A\alpha }\partial R_{B\beta }$. The phonon spectrum is the spectrum of the BO surface’s Hessian — full stop. A soft mode is a direction of vanishing curvature; an imaginary mode is a direction of negative curvature, i.e. the high-symmetry $\mathbf{R}$ is a saddle, not a minimum, of the surface.

Stating it plainly: the surface defined here is the precondition for the entire phonon program — the dynamical matrix is literally its mass-weighted Hessian, and anharmonicity is just the cubic and higher terms of its Taylor expansion.

6. Where Born–Oppenheimer breaks (limit of validity)

BO is an approximation with a well-defined failure regime, and the same Born–Huang expansion that justifies it also exhibits exactly the terms it discards. Writing the exact wavefunction as $\Psi ={\sum }_n{\chi }_n(\mathbf{R}){\psi }_{\text{elec}}^{(n)}(\mathbf{r};\mathbf{R})$ and projecting the full $H$ onto the electronic states, $T_n$ acting on the $\mathbf{R}$-dependent electronic states generates non-adiabatic (derivative) coupling terms between surfaces,

$${\mathbf{d}}_{ij}(\mathbf{R})=⟨{\psi }_{\text{elec}}^{(i)}\left|{\nabla }_{\mathbf{R}}\right|{\psi }_{\text{elec}}^{(j)}⟩,$$

which the (adiabatic) BO approximation simply drops — it keeps only the diagonal surface and the diagonal Born–Huang correction. First-order perturbation theory shows the coupling scales as

$${\mathbf{d}}_{ij}\sim \frac{⟨{\psi }^{(i)}|{\nabla }_{\mathbf{R}}{H}_{\text{elec}}|{\psi }^{(j)}⟩}{E_{\text{elec}}^{(j)}-E_{\text{elec}}^{(i)}},$$

so it is controlled by the electronic energy gap in the denominator. The approximation is excellent when surfaces are well separated and degrades precisely as that gap closes:

• Conical intersections. Where two surfaces become degenerate, $E^{(i)}(\mathbf{R})=E^{(j)}(\mathbf{R})$, the denominator vanishes and ${\mathbf{d}}_{ij}$ diverges. Generically (for a real Hamiltonian) the crossing seam has codimension two, so the surfaces form a double cone — a conical intersection — and BO is qualitatively wrong in its vicinity: the electronic character switches over a tiny region of $\mathbf{R}$ and a Berry phase of $\pi$ is acquired on encircling the seam. These are not pathologies of small molecules only; they govern photochemistry, internal conversion, and radiationless decay (Yarkony).
• Near-degeneracy / small gaps. Even without an exact crossing, a small but finite gap makes ${\mathbf{d}}_{ij}$ large. This is the regime of avoided crossings, vibronic coupling, the Jahn–Teller and pseudo-Jahn–Teller effects, and any open-shell or near-metallic system where excited states lie within a vibrational quantum of the ground state.
• Strong electron–phonon coupling. When the lattice distortion moves electronic states by an amount comparable to the electronic energy scale, the parametric/adiabatic factorization itself is questioned — polaron formation, Peierls and charge-density-wave instabilities, and metals (gapless by construction, so there is no protective gap at the Fermi level) all push on BO. In metals the surface picture survives for the nuclei but the adiabatic separation of individual electronic states does not, which is why phonon-induced superconductivity and finite phonon linewidths live in the corrections, not the BO surface itself.

The honest summary: BO gives a single, well-defined surface $E^{(n)}(\mathbf{R})$ that is the correct zeroth-order object whenever the electronic gap is large compared with nuclear energy scales, and it fails — by construction, through the diverging ${\mathbf{d}}_{ij}$ — exactly where that gap closes. Below that everything else here is built; above it, one needs the full Born–Huang coupled-channel treatment.

Builds toward: dynamical matrix & ASR · Hellmann–Feynman forces · Kohn–Sham DFT

Key references

• Original adiabatic separation — M. Born & J. R. Oppenheimer, Ann. Phys. 389, 457 (1927).
• Rigorous $\kappa =(m_e/M{)}^{1/4}$ expansion — M. Born & K. Huang, Dynamical Theory of Crystal Lattices (Oxford Univ. Press, 1954), App. VIII.
• Conical intersections & non-adiabatic coupling — D. R. Yarkony, Rev. Mod. Phys. 68, 985 (1996).