The puzzle guide 2 left behind
In guide 2 you built up the grain boundary from scratch. A low-angle boundary you understood exactly — it is just a tidy wall of dislocations, and its energy climbs smoothly as you crank the misorientation up. But once the misorientation passes roughly 15 degrees the dislocation cores overlap and the picture dissolves into a high-angle boundary, a genuinely disordered seam. The surprise was the energy curve. Across almost the whole high-angle range the grain-boundary energy barely moves — it sits on a broad, nearly flat plateau around 0.3 to 1 joule per square metre, as if the crystal simply does not care exactly how the two grains are twisted.
Almost. If you measure carefully — and people have, for decades — the plateau is not perfectly flat. At a handful of very specific misorientations the energy plunges into a sharp, narrow dip, a cusp, like a single deep pothole in an otherwise level road. Rotate two copper grains 60 degrees about a body-diagonal direction and the boundary between them can cost a tenth of its neighbours. Rotate them a couple of degrees away from that magic angle and the energy leaps back up to the plateau. Something about a few exact orientations is special, and nothing in the ordinary picture of a disordered high-angle boundary explains it.
Overlay two grids and rotate: the coincidence-site lattice
Here is the idea that cracks it open, and you can do it on a light table. Take two identical sheets of transparent graph paper — two copies of the same crystal lattice — lay one exactly on top of the other so every point coincides, then pin them at one shared point and rotate the top sheet. At a general angle you get a moire pattern, a shimmering mess in which almost no dot of the top grid lands squarely on a dot of the bottom grid. Away from that one pinned point, the two lattices are hopelessly out of registry. This is the visual truth of a random high-angle boundary: two crystals sharing a seam but matching almost nowhere.
Now rotate slowly and watch. At most angles the mess just churns. But at a few precise angles something clicks: suddenly a whole regular grid of points snaps into perfect overlap, dots of the top sheet sitting dead-centre on dots of the bottom sheet, spread evenly across the entire sheet — not just at the pin. Those shared points, the ones common to BOTH rotated lattices at once, form a new, coarser lattice of their own. That is the coincidence-site lattice, or CSL. It is a sublattice — a lattice built from a subset of the original points — and it exists only at these special misorientations, which is exactly why only special misorientations behave differently.
Do it once with real numbers in a flat square lattice, because it is the cleanest case. Rotate the top grid by an angle whose cosine is 4/5 — that is 36.87 degrees — about the axis sticking out of the page. Run the geometry (it is just the 3-4-5 right triangle in disguise) and you find that exactly one point in every five now coincides between the two grids. The coincidence points sit on a square lattice five times larger in area than the original cell. One in five: the density of coincidence sites is 1/5, and that reciprocal — the number 5 — is the whole story of the next section.
The Sigma value: counting the coincidences
The number that names a special boundary is the Sigma value, written the Greek letter Sigma. It is defined with disarming simplicity: Sigma is the reciprocal of the fraction of lattice points that coincide. If one point in five is shared, Sigma equals 5. If one in three, Sigma equals 3. Equivalently — and this is the version to picture — Sigma is the volume of the CSL unit cell divided by the volume of the ordinary crystal unit cell. A big Sigma means coincidences are rare and far apart (a large, sparse CSL); a small Sigma means coincidences are dense and frequent (a tight CSL). Small Sigma is what we are hunting, because it means the two crystals share a lot of points.
COMMON COINCIDENCE BOUNDARIES (cubic crystals)
Sigma misorientation note
----- ---------------------- ---------------------------
1 0 deg (any axis) perfect crystal, no boundary
3 60.0 deg about <111> the COHERENT TWIN (guide 4)
5 36.9 deg about <100> the classic CSL example
7 38.2 deg about <111>
9 38.9 deg about <110>
11 50.5 deg about <110>
Sigma = volume(CSL cell) / volume(crystal cell)
= 1 / (fraction of lattice points that coincide)
In a CUBIC lattice, Sigma is ALWAYS an ODD number.
(Sigma 1 = a perfect single crystal: every point coincides.)Two honest facts about that table. First, Sigma is always odd in a cubic crystal — never 2, 4, or 6. That is not a convention or a rounding; it falls straight out of the number theory of the rotation matrices that produce coincidence, and any even factor can always be divided back out. Second, Sigma 1 is the special boundary that is no boundary at all: zero misorientation, every single point coincides, a perfect single crystal. Reading the table you can feel the ladder — Sigma 3 shares a third of all points, Sigma 5 a fifth, Sigma 11 only a eleventh — and you can already guess that the low-Sigma boundaries near the top are the ones that will turn out cheap.
Why coincidence buys low energy
The link from geometry to energy is the whole point, and it is physical, not mystical. Recall from guide 1 why any interface costs energy at all: atoms at the seam sit in a wrong neighbourhood, their bonds stretched, bent, or broken, and every strained bond is stored energy. Now suppose the boundary plane is chosen to run through the coincidence-site lattice — through the points the two crystals share. At each of those shared points an atom fits BOTH crystals perfectly at once; it is a happy atom, in good registry with both sides, its bonds close to ideal. A boundary threaded with coincidence sites therefore has far fewer badly-strained bonds than a random seam, and fewer strained bonds means lower grain-boundary energy. That is the cusp.
But real boundaries never sit at the exact magic angle — nature is sloppy, grains grow where they grow. So how far off can you be and still count as special? The rule of thumb is the Brandon criterion: a boundary still behaves as a Sigma boundary if its misorientation lies within about 15 degrees divided by the square root of Sigma. For Sigma 3 that is a generous 8.7 degrees of slack; for Sigma 5, about 6.7 degrees; for a high Sigma 25, only 3 degrees. The crystal absorbs that small leftover mismatch by sprinkling a sparse array of localised grain-boundary dislocations into the otherwise-perfect coincidence pattern — the exact same trick a low-angle boundary used, now laid on top of the CSL instead of on top of a perfect crystal. Special boundaries, in other words, tolerate a little imperfection gracefully.
The champion: Sigma 3, the coherent twin
One special boundary towers above all the rest, and it is the smallest Sigma there is (other than the no-boundary Sigma 1): the Sigma 3 coherent twin boundary. A third of all lattice points coincide across it — the densest possible sharing for a real boundary — and when the boundary plane is laid exactly on the mirror plane, the two grains are perfect mirror images of each other. Look across the seam and it is as if the crystal reflected in a flawless mirror: not a single bond is broken, and only the second-neighbour stacking differs. In an FCC metal this is nothing more than one hiccup in the ABCABC stacking you met in the close-packing rung — a layer laid down as if reflected, ABC|CBA.
The energy payoff is spectacular and worth a real number. A random high-angle boundary in copper costs roughly 0.6 joule per square metre. The coherent Sigma 3 twin in the same copper costs only about 0.02 to 0.04 — fifteen to thirty times cheaper. That enormous discount is why annealing twins are everywhere in copper, brass, and austenitic steels: when these crystals grow, throwing down a nearly-free twin boundary barely costs anything, so they do it constantly, and a polished section is laced with the straight, parallel-sided bands twins leave behind. You will meet twins properly in the very next guide — how they form, coherent versus incoherent, and twinning as a deformation mechanism. Here the point is narrower and sharper: the coherent twin is simply the extreme, champion case of the CSL story, the boundary where coincidence is densest and energy is lowest.
Why any of this matters: energy runs the microstructure
Grain-boundary energy is not a curiosity — it is a driving force, and it quietly steers the whole orientation geography of a real metal. A grain boundary is stored energy per unit area, so a polycrystal with a lot of boundary area is like a stretched spring: the system can lower its free energy simply by having less boundary. That is exactly what grain growth does. On heating, boundaries migrate so that big grains eat small ones, the total boundary area shrinks, and the average grain size creeps up — the microstructure coarsening to pay off its interfacial-energy debt, curvature by curvature. The same accounting explains wetting: a high-energy boundary can be worth replacing with two lower-energy ones if a second phase creeps in along it.
This is where the special boundaries stop being a geometric curiosity and start earning their keep. Low-Sigma boundaries are not only low in energy; they are also, as a rule, more sluggish to migrate, more resistant to sliding, and far harder for corrosion or cracks to penetrate along. So an engineer can ask: could we deliberately give a metal MORE special boundaries and fewer nasty random ones? That is grain-boundary engineering — thermomechanical processing that raises the fraction of Sigma 3 (and Sigma 9, Sigma 27) twin-related boundaries — and in nickel alloys and stainless steels it measurably improves resistance to cracking and corrosion. The whole strategy rests on the CSL idea you just built: some seams, by pure geometric luck, are simply better neighbours.
Where this rung goes from here. You now have the whole vocabulary of special same-phase boundaries. Guide 4 zooms into the champion — twins, and their close cousins the stacking fault and the antiphase boundary in an ordered alloy — as planar mistakes in the stacking. Guide 5 then crosses the biggest divide of all: the interphase boundary between two DIFFERENT phases, where the two lattices need not even match, and coincidence gives way to coherent, semicoherent, and incoherent matching, misfit dislocations, and epitaxy. The CSL you learned here is the same-phase warm-up for that harder, richer problem.