The mismatched seam where two grains meet
Almost every metal, ceramic, and rock you have ever touched is a polycrystal: not one crystal but a mosaic of thousands of small ones, each a little grain with its own orientation, packed together to fill space. Where two grains butt against each other, their atomic patterns almost never line up — imagine floor tiles of the same design laid down in two patches rotated at different angles, so along the join the rows meet at a mismatched seam. That internal seam is a grain boundary: a two-dimensional, planar defect separating two crystals of the very same phase that differ only in orientation. It is not a crack and not a different material; it is the same crystal on both sides, just turned.
In guide 1 you met the most extreme planar defect of all — the free surface, where the crystal simply ends and half of every atom's bonds dangle into empty space, costing a large surface energy. A grain boundary is gentler. The atoms in the seam still have neighbours on both sides; their bonds are stretched, squeezed, and bent out of the ideal geometry, but they are not cut. So a boundary costs energy — the pattern is distorted where the two grains fail to register — but far less than a free surface. As a rule of thumb a grain boundary in a metal costs roughly a third of the surface energy: a copper free surface is about 1.7 J/m^2, while a general copper grain boundary is nearer 0.5 J/m^2.
A small-angle boundary is literally a wall of dislocations
Here is the single most beautiful idea in this guide. When the misorientation is small — a few degrees — a low-angle grain boundary is not a smeared-out mess at all. It is a neat, periodic array of dislocations, the same line defects you studied in the previous rung. Take two identical crystals and tip one by a small angle about an axis lying in the plane where they meet; this is a tilt boundary. Nearly every atomic row from the left grain finds a partner in the right grain and joins up smoothly. But every so often a row on one side has no partner, so an extra half-plane of atoms has to be wedged in to take up the slack — and an extra half-plane is exactly an edge dislocation. A small-angle tilt boundary is therefore a vertical stack of edge dislocations, evenly spaced.
The spacing follows from pure geometry. To close an angular gap of theta you need one extra half-plane every so often, and the spacing between neighbouring dislocations comes out as D = b / (2 sin(theta/2)), which for small angles is just D = b / theta, with b the length of the Burgers vector. Put numbers on it for copper, where b is about 2.5 angstrom. At theta = 1 degree (that is 0.0175 radians) the dislocations sit D = 2.5 / 0.0175 = 143 angstrom apart, roughly 14 nm — comfortably far, with tens of near-perfect atomic rows between each one. The strain is bottled up in the isolated dislocation cores, and the crystal in between barely notices.
SYMMETRIC TILT BOUNDARY = a wall of edge dislocations
grain A (tipped +theta/2) | grain B (tipped -theta/2)
\ \ \ \ | / / / /
\ \ \ \ _|_ <- extra half-plane /
\ \ \ \ | = an EDGE dislocation
\ \ \ \ | / / / /
\ \ \ \ _|_ / / / /
\ \ \ \ | / /
\ \ \ \ | / / / /
\ \ \ \ _|_ / / / /
|
spacing of the _|_ 's : D = b / (2 sin(theta/2)) ~ b/theta
Cu, b ~ 2.5 A :
theta = 1 deg -> D ~ 143 A ~ 14 nm (dislocations far apart)
theta = 10 deg -> D ~ 14 A ~ 1.4 nm (cores almost touching)
TILT : rotation axis lies IN the boundary -> wall of EDGE dislocations
TWIST : rotation axis is NORMAL to boundary -> grid of SCREW dislocationsTilt is only half the family. If instead you rotate one grain about an axis perpendicular to the boundary plane, the rows twist past one another like two combs turned face to face, and the mismatch is taken up not by edge dislocations but by a crossed grid of screw dislocations — a twist boundary. Real boundaries are usually a mix of tilt and twist, just as a real dislocation is a mix of edge and screw. The headline is the same for both: at small angles, a grain boundary is nothing more exotic than an orderly wall or grid of the dislocations you already understand.
Cranking up the angle: the low-to-high crossover
Watch what happens as you keep turning the grains apart. The spacing D = b/theta shrinks: double the angle and you halve the distance between dislocations. Look at the second number in the sketch — at theta = 10 degrees the copper dislocations are only about 14 angstrom apart, a mere five or six atoms. Their strained cores are now practically touching, overlapping and merging. At that point it stops making sense to speak of individual, countable dislocations at all: the boundary has become a continuous ribbon of badly-fitting atoms, and we call it a high-angle grain boundary. The crossover happens gradually, typically somewhere around 10 to 15 degrees — an honest rule of thumb, not a sharp line in nature.
The energy tells the same story quantitatively. Summing up the strain fields of a wall of dislocations gives the Read-Shockley formula for the grain-boundary energy: gamma = gamma_0 times theta times (A - ln theta), where gamma_0 and A are constants set by the material. Do not memorise it; feel its shape. At tiny theta the boundary is nearly free of dislocations, so its energy is small — but because each new dislocation you add carries its own line energy (which scales as b^2), the energy climbs steeply at first, thanks to that -ln theta term. As theta grows and the cores begin to overlap, adding more dislocations buys you less and less, so the curve bends over and flattens into a plateau. That plateau is the roughly-constant energy of a general high-angle boundary.
- Take two identical crystals; rotate one by a small angle theta about an axis lying in the boundary plane.
- Along the join most atomic rows match up, but every so often one is left unpartnered, forcing in an extra half-plane — one edge dislocation.
- Geometry fixes their spacing as D = b / (2 sin(theta/2)) ~ b/theta, so bigger theta means dislocations packed closer together.
- Add up the dislocations' strain fields to get the boundary energy: the Read-Shockley curve, rising steeply then levelling off.
- Past roughly 15 degrees the cores overlap and merge; individual dislocations lose meaning and you have a high-angle boundary.
Not all high-angle boundaries are equal
It would be tidy if every high-angle boundary sat at the same plateau energy, but nature is more interesting than that. Plot boundary energy against misorientation angle and the plateau is not flat — it is dotted with sharp downward dips, called cusps, at particular special misorientations. At those angles the two rotated lattices, though generally out of step, happen to share a regular sub-pattern of atomic sites that fall on both grids at once. Where atoms from both grains can sit on shared positions, the boundary fits together far better and costs much less energy. This shared sub-pattern is the coincidence-site lattice, and it is the whole subject of the next guide.
The deepest cusp of all is a boundary so well-matched that it barely costs anything: the twin boundary, where the crystal on one side is an exact mirror image of the crystal on the other. Every atom in the boundary plane sits in a perfectly shared position, bonds are almost undistorted, and the energy can be an order of magnitude below a general boundary — a coherent twin in copper costs only about 0.02 J/m^2, against 0.5 J/m^2 for a random boundary. Twins are why annealed brass and bronze are threaded with those straight, parallel bands under the microscope. We will meet twinning, and its cousins the stacking fault and antiphase boundary, in guide 4.
Why it matters: energy that drives grains and liquids
A grain boundary is stored energy, and a material is always looking for ways to shed it. The simplest way is to have less boundary — fewer, bigger grains means less total boundary area to pay for. That is exactly what happens when you heat a fine-grained metal: the boundaries migrate, small grains are eaten by their larger neighbours, and the average grain size coarsens. This is grain growth, and its engine is boundary curvature. A curved boundary feels a pressure pushing it toward its own centre of curvature, because moving that way shrinks its area; so big grains with gently bulging boundaries swallow small grains with tightly-curved ones, and the tiniest grains vanish altogether.
The same energy balance sets the angles where boundaries meet. Where three grains come together along a line, the three boundaries pull on that line like three ropes, each with a tension equal to its energy per unit area. If all three boundaries have the same energy the ropes balance at 120 degrees apart — which is why an equilibrium grain structure looks like a honeycomb of roughly hexagonal cells. Deviations from 120 degrees are a direct readout of which boundaries are low-energy special ones and which are ordinary.
Push this a little further and you get wetting, one of the most consequential facts in metallurgy. Suppose a thin film of a second phase — a liquid, or a brittle impurity — could sit inside a grain boundary. Whether it spreads along the boundary or beads up is decided by a tug-of-war between the boundary's energy and the energy of the two new interfaces the film would create. When the grain-boundary energy exceeds twice the film-to-grain interfacial energy, the film wins and creeps all the way along every boundary, coating each grain completely. This is exactly how a trace of low-melting metal can make a whole alloy crumble, and why controlling grain-boundary chemistry can be a matter of life and death for an engineering part. The same interfacial-energy accounting will govern the interphase boundaries of the final guide.