Periodic Electronic-Structure Calculations

Introduction

A crystal contains an immense number of atoms, yet a periodic calculation can model it through a small repeating cell. The price of that efficiency is a set of assumptions: perfect repetition, chosen cell geometry and finite sampling in reciprocal space. A solid's energy, forces and band structure can change when the plane-wave basis or k-point mesh is too coarse. Understanding these controls prevents a polished electronic band plot from being mistaken for a converged material prediction.

Core explanation

The unit cell is defined by lattice vectors and atomic positions within them. Periodic boundary conditions repeat the cell indefinitely, so an atom near one face interacts with the image of atoms across that face. A primitive cell contains the smallest translational repeat; a conventional cell may be larger to display symmetry more clearly. Defects, surfaces and dilute dopants require larger supercells because repeating a small defect cell creates an unrealistically dense array of identical defects. A molecule placed in a periodic code is likewise repeated, so sufficient vacuum and attention to image interactions are needed when isolated behavior is intended.

For an ideal crystal with a periodic potential, Bloch's theorem permits electronic states labeled by a band index n and wave vector k. The wave vector is considered within the first Brillouin zone of reciprocal space. At each k, the cell-periodic part of the state can be expanded in plane waves whose reciprocal vectors G satisfy a cutoff condition. Raising the plane-wave kinetic-energy cutoff adds basis functions and usually improves representation of rapidly varying features, but increases cost. A pseudopotential or projector-augmented-wave treatment commonly handles difficult core-region behavior; its recommended cutoff is a starting point, not proof of convergence. VASP's developer documentation on plane waves and periodicity gives the Bloch and cutoff construction.

The Brillouin-zone integral is approximated by a finite grid of k points with weights. A coarse mesh may be adequate for a large real-space cell or a wide-gap insulating system, while a metal's partially occupied bands can demand denser sampling. k-point sampling and plane-wave cutoff are distinct numerical axes: increasing one does not repair error in the other. VASP's k-point documentation explicitly treats sampling convergence as essential, and its integration guidance explains symmetry reduction. Occupation smearing for metals can aid integration and convergence, but its width affects energies and should be tested for the intended quantity.

A band structure plots selected eigenvalues versus k along a chosen reciprocal-space path. It can reveal approximate dispersions, band crossings and character, but the path alone is not a complete Brillouin-zone integration. A density of states uses k-point information over a grid and chosen broadening. The electronic gap inferred from approximate Kohn–Sham bands is not automatically an experimental optical or fundamental gap. For structural properties, compare total energies and forces rather than relying on orbital plots. A high-symmetry path can miss a band extremum located elsewhere, so gap classification may require more complete sampling.

Convergence tests should target differences and observables, not only a large absolute total energy. If comparing two crystal polymorphs, compute both with compatible cell choices, potential libraries, cutoffs and k-point density, then test whether their energy difference and relative ordering stabilize. A force or stress calculation may need a stricter cutoff than a qualitative density plot. Variable-cell optimization can be affected by finite-basis stress; a reported lattice parameter should be checked against cutoff and k-point changes. Magnetic order, defects and surface coverage introduce additional cell-size checks. Numerical convergence cannot remove exchange–correlation functional error, and a higher cutoff cannot fix an inappropriate structural model.

Step-by-step reasoning

1. Select a chemically justified cell, lattice vectors and atomic arrangement; decide whether a supercell is required. 2. Choose a compatible electronic method and core treatment for every element. 3. Converge plane-wave cutoff while holding the k-point scheme sufficiently accurate. 4. Converge k-point sampling and, for a metal, occupation smearing for the target energy, force or property. 5. Optimize structure only after numerical settings are adequate; repeat relevant checks if cell dimensions change markedly. 6. Interpret bands or DOS with energy zero, k path, projections and functional limits documented.

Visual explanation

Draw a square unit cell tiled across a plane, with one atom near its right edge and an image atom across the next cell. Beside it draw the reciprocal-space Brillouin zone as a square dotted with a mesh of k points. A separate circle of G vectors inside a cutoff radius represents the plane-wave basis at one k. Increasing the dotted mesh changes sampling; enlarging the G circle changes basis size. The two pictures make clear why both convergence tests are needed.

Real-world analogy

A repeating wallpaper design can be described by one tile, but a stain added to that tile repeats everywhere unless a larger pattern is used. A periodic supercell similarly repeats a modeled defect. Meanwhile, sampling k points is like checking many viewing angles of the repeated pattern; expanding plane waves is like increasing image resolution at each angle. More resolution at one angle does not replace the missing angles.

Real-world example

A materials group compares two candidate crystal structures of a battery electrode. A quick calculation suggests one is lower by 2 meV per formula unit, a tiny margin. The group increases cutoff and k-point density and discovers that the energy difference changes sign under the first coarse mesh. It then uses consistent core data, checks magnetic order and relaxes both structures at converged settings. The final claim includes numerical convergence and functional uncertainty. A pretty band structure from the first run would have been insufficient evidence for phase stability.

Why?

Why can a larger supercell reduce artificial defect interactions? Under periodic boundaries, every defect is copied with the cell. Increasing the cell separates those copies and better approximates an isolated defect, although long-range electrostatic and elastic effects may converge slowly. This is a model-size issue distinct from adding more plane waves or k points to a fixed cell. A cell-size test is therefore necessary when a dilute limit is the target.

Common misconception

“A unit-cell calculation includes all real crystal disorder.” It represents ideal repetition unless disorder is modeled explicitly. “More plane waves guarantee a converged total energy difference.” k-point and cell-size errors can remain. “Gamma-only sampling is always enough for a large-looking molecule or cell.” The required mesh depends on reciprocal-space variation and target precision. “A DFT band gap is automatically an optical absorption threshold” ignores functional and excited-state effects.

Worked example

Two hypothetical phases A and B have energies per formula unit of −100.000 and −99.997 eV at a low cutoff, so A appears lower by 3 meV. At a higher cutoff their energies are −100.030 and −100.032 eV, so B appears lower by 2 meV. The absolute energy shift of roughly 30 meV is less important than the changed ordering of the 2–3 meV difference. Now suppose a denser k grid changes B's relative energy by another 1 meV; the phase conclusion remains delicate. A responsible report gives the converged energy difference, its remaining numerical spread and the functional used, not just the first sign. These invented numbers illustrate sensitivity, not a particular material.

Quick check

1. What does raising the plane-wave energy cutoff chiefly increase? Answer: The number of plane-wave basis functions available at each sampled k point. 2. What does increasing the k-point mesh chiefly improve? Answer: Numerical sampling of the Brillouin-zone integrals.

Exam focus

Describe how lattice vectors and a motif create an ideal periodic crystal. Identify reciprocal-space k points and the plane-wave cutoff as different convergence controls. Explain why a defect in a small cell is an array of defects. Given an energy table, compare differences under increasing cutoff and k density. State why a computed band gap or DOS needs method and sampling context before experimental interpretation.

Advanced insight

The reciprocal grid needed for a metal is tied to its Fermi surface and chosen smearing; a semiconducting cell of similar size may converge differently. Symmetry can reduce the number of irreducible k points without changing the underlying full-grid integral, but a defect or magnetic order can remove that symmetry. Stress convergence can be more demanding than energy convergence during variable-cell relaxation. Charged defects create additional electrostatic image interactions that may require specialized corrections beyond simply enlarging the cell. Every numerical test should be linked to the final observable and accuracy target.

Summary

Periodic electronic-structure calculations use a repeating unit cell and reciprocal-space Bloch states to represent solids efficiently. Plane-wave cutoff controls basis flexibility, while k-point sampling controls Brillouin-zone integration; cell size controls repeated-image artifacts. All three may need convergence checks, depending on the question. Bands and DOS are useful model outputs, but stable material conclusions require consistent total-energy or property comparisons and explicit limits of the structural and electronic model.

Practice questions

1. Can doubling the plane-wave cutoff compensate for an inadequate k-point grid in a metal? Answer: No. They control different numerical approximations. 2. Why can a small defect supercell misrepresent a dilute impurity? Answer: Periodic repetition places identical impurities too close together, allowing artificial interactions. 3. Does a band plot along one high-symmetry path necessarily locate the true minimum gap? Answer: No. A band extremum may occur away from the plotted path. 4. Two phases differ by 1 meV per formula unit, but the cutoff test changes their difference by 4 meV. Is their order numerically established? Answer: No. The numerical variation exceeds the claimed separation.