Grain Growth: How Metals Get Their Microstructure

A metal is a mosaic of crystals

Almost no engineering metal is a single crystal. When molten metal solidifies, or when a heavily deformed metal recrystallizes, it does not freeze into one continuous lattice — it nucleates as thousands of tiny crystals, each with its own random lattice orientation, growing outward until they collide with their neighbours. The resulting patchwork of differently oriented crystalline regions is the metal's grain structure, and the boundaries between grains — thin, disordered transition zones only a few atoms wide — are called grain boundaries. This microstructure, not the underlying crystal lattice itself, is what a metallurgist looking through a microscope actually sees.

Nucleation gives a Voronoi tessellation

If nuclei form at random locations and each grows outward at a uniform rate until it meets its neighbours, the resulting cell pattern is geometrically a Voronoi tessellation — every point in space belongs to whichever nucleus is closest, and the grain boundaries are exactly the equidistant lines between neighbouring nuclei. This is the same construction used to model everything from cellular biology to territory maps, and it is a reasonable first approximation for a freshly solidified or freshly recrystallized metal before any subsequent boundary motion has had time to occur.

Annealing sets boundaries in motion

A freshly formed grain boundary carries excess energy per unit area — the atoms along it sit in a disordered, higher-energy configuration than the ordered lattice on either side. Heat the metal (anneal it) and that boundary energy becomes mobile: a curved boundary migrates toward its own centre of curvature, exactly the way a soap film shrinks to minimize its surface area, because doing so reduces the total boundary energy of the system. A perfectly flat boundary between two grains does not move; only curvature drives migration.

The von Neumann-Mullins law

In a 2D polycrystal, curvature-driven migration produces a strikingly simple topological result, derived independently by John von Neumann and William Mullins:

dA/dt = (πMγ/3) · (n - 6)

A = grain area,  n = number of sides (neighbouring grains),
M = boundary mobility,  γ = boundary energy per unit length

A grain's area growth rate depends only on how many sides it has, not on its actual size or shape. Grains with fewer than six sides (n < 6) have net-concave boundaries on average and shrink; grains with more than six sides (n > 6) are net-convex and grow; a hexagonal grain (n = 6) is in instantaneous equilibrium. As small, low-side-count grains vanish and are absorbed by their larger, many-sided neighbours, the average number of sides across the whole microstructure drifts toward six — which is why real annealed microstructures are dominated by roughly hexagonal-looking grains, echoing the same geometric optimum that appears in honeycomb and soap-froth structures.

Coarsening and grain-size distribution

Repeated over the whole microstructure, this process is called grain growth or coarsening: the total number of grains steadily decreases and the average grain size increases, following approximately d² - d₀² ∝ t (a parabolic growth law), because larger average grain size means, on average, larger radii of curvature and therefore slower migration — growth naturally decelerates as the microstructure coarsens, though it never fully stops as long as any curvature remains.

Why grain size sets a metal's strength: Hall-Petch

Grain boundaries are not just a cosmetic detail — they are the single biggest lever engineers have over a metal's mechanical strength. Plastic deformation happens when line defects called dislocations glide through the crystal lattice, and a grain boundary is a barrier a dislocation cannot easily cross, because the lattice orientation changes abruptly there. The Hall-Petch relationship, developed independently by E. O. Hall and N. J. Petch in the 1950s, quantifies the result:

σ_y = σ₀ + k / √d

σ_y = yield strength,  σ₀ = friction stress inside a grain,
k    = a material-specific constant,  d = average grain diameter

Smaller grains mean more boundaries per unit volume, and each dislocation travels a shorter average distance before hitting one and piling up — so a fine-grained metal needs more applied stress to yield than a coarse-grained one of identical composition. This is why controlling annealing time and temperature (which controls how far grain growth is allowed to proceed) is one of the primary tools in metallurgical processing, and why the same alloy can be engineered to be significantly stronger or softer purely by controlling its grain size.

Frequently asked questions

Try it live

Everything above runs in your browser — open Grain Growth and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.