The crystal simply stops: the free surface
The last two rungs walked down a ladder of dimension. First the point defect — a single wrong site, zero-dimensional, a lone hole in the wallpaper. Then the dislocation — a one-dimensional line threading the crystal, the ruck in the rug that lets metals bend. This rung takes the next step up: the two-dimensional, or planar, defect — a whole sheet across which the structure changes. And the simplest, most drastic planar defect of all is the one where the crystal does not meet another crystal but meets nothing at all: the free surface, the outer skin where the lattice simply stops and vacuum begins.
Picture the infinite wallpaper of the crystal sliced clean in half. Every atom in the bulk sat snug in its full cage of neighbours — in a close-packed metal, twelve of them, coordination 12. But an atom left sitting on the fresh surface has lost the neighbours that used to sit above it. Those upward bonds now reach out into empty space, unsatisfied — dangling bonds. Every dangling bond is a scrap of bonding energy left unpaid, so the surface atoms sit at higher energy than the ones buried inside. That excess energy, counted per unit of surface area, is the surface energy, written gamma, in units of joules per square metre (J/m^2) — the price, per square metre, of having a surface at all.
Counting the cost: broken bonds and surface energy
Let us put a number on it with a tiny broken-bond estimate. In copper the cohesive energy is about 3.5 eV per atom, spread over 12 bonds, so each bond is worth roughly 2 x 3.5 / 12 = 0.6 eV (each bond shared between two atoms). Cleave along a close-packed {111} plane and each surface atom loses 3 of its 12 neighbours (its coordination drops from 12 to 9). Cleaving makes two surfaces at once, so split the broken-bond energy between them: about (1/2) x 3 x 0.6 = 0.9 eV per surface atom. Pack in the {111} surface atom density, about 1.8 x 10^19 atoms per m^2, and gamma = 0.9 eV x (1.6 x 10^-19 J/eV) x (1.8 x 10^19 /m^2), which comes to roughly 2.6 J/m^2.
The measured surface energy of copper is about 1.8 J/m^2, so our back-of-envelope count lands in the right ballpark but runs high, by roughly 40 percent. That gap is honest and instructive: the simple model freezes the atoms in place with naked, unsatisfied bonds, but real surfaces relax and their electrons redistribute to heal some of the damage — exactly the subject of the next section. Typical metals sit around 1 to 3 J/m^2; oxides and more weakly bound solids are lower. In per-atom terms these are fractions of an eV per surface atom — a real cost, but only a slice of a full bond.
Notice that the cost depended on WHICH plane we cut. That is the deep point: surface energy is anisotropic — it depends on orientation. A close-packed {111} face loses only 3 bonds per atom; a {100} face loses 4 (coordination 8), and more open planes lose still more. So gamma for {111} is less than for {100}, which is less than for {110}, and a crystal left to find its lowest-energy shape will grow flat facets on its cheap close-packed planes rather than round itself off. That is why well-formed crystals, and even tiny nanoparticles, are faceted rather than spherical: the equilibrium shape is the one that minimises the total surface energy summed over all its faces.
The surface fights back: relaxation and reconstruction
Those dangling bonds are expensive, and atoms are not obliged to sit still and pay full price. The gentlest response is surface relaxation: the whole outermost layer shifts slightly, almost always sinking inward so the spacing between the top layer and the one beneath it shrinks by a few percent. Crucially, the atoms keep their sideways positions — the two-dimensional pattern of the surface is exactly the truncated bulk, just with a squeezed top interlayer spacing. One honest subtlety for solids: the energy to CREATE new surface area (the gamma we have been counting) is not quite the same as the surface STRESS, the work to elastically stretch an already-existing surface — for a liquid the two coincide as a single surface tension, but for a solid they differ.
The bolder response is surface reconstruction: the surface atoms rearrange sideways into a brand-new two-dimensional periodicity, different from a plain slice of the bulk, so as to pair up or eliminate dangling bonds. The classic case is silicon. On the Si(100) surface each atom would dangle two bonds; instead neighbouring atoms tilt toward each other and bond into dimers, and the rows of dimers double the surface repeat to a (2x1) cell — halving the dangling-bond count at a stroke. Even more spectacular is Si(111), which reconstructs into a huge (7x7) cell, its surface unit cell 7 times the bulk spacing in each direction; gold famously reconstructs its (111) face into a herringbone pattern. The notation (m x n) just says how many bulk surface cells the new surface cell spans.
The rule of thumb: the more directional and strong the dangling bonds, the more the surface wants to reconstruct. Covalent semiconductors, whose broken bonds are sharp, directional, and costly, reconstruct dramatically; simple metals, whose electron sea heals over a cut more smoothly, mostly just relax. Either way, these are the CLEAN-surface structures — a real surface exposed to air is instantly buried under adsorbed gas and oxide, so seeing a bare (7x7) at all demands ultra-high vacuum. When surface scientists first imaged the Si(111)-(7x7) directly with the scanning tunnelling microscope, atom by atom, it was one of the great confirmations that the atomic picture of surfaces is real.
SIDE VIEW THROUGH A CRYSTAL SURFACE (o = atom, | - = bonds, : = dangling bond)
(a) IDEAL TRUNCATION (b) RELAXATION (c) RECONSTRUCTION
naked dangling bonds top layer sinks in top atoms re-pair (dimers)
: : : : <-vacuum : : : : o-o o-o <-dimers
o o o o surface o o o o (d smaller) o o o o
| | | | | | | | | | | |
o--o--o--o layer 2 o--o--o--o o--o--o--o
| | | | | | | | | | | |
o--o--o--o bulk (12) o--o--o--o o--o--o--o
same 2D cell same 2D cell, smaller d NEW, larger 2D cellWhy surfaces run the show
For a big crystal, surface energy is a rounding error: only a sliver of atoms live on the skin. But shrink the crystal and the balance tips, because the surface-to-volume ratio climbs. Halve a particle radius and the fraction of atoms at the surface roughly doubles; in a 3 nm nanocrystal a large share of every atom is a surface atom. Now that stored surface energy is a major part of the whole particle energy, and it starts to change the material outright: nanoparticles show melting-point depression, melting tens or hundreds of degrees below the bulk, because their surface atoms, already half-unbonded, come loose more easily. Those same under-coordinated, reactive surface atoms are why finely divided metals catalyse reactions and why some nanopowders are pyrophoric.
From the outer skin to internal seams: the rest of the rung
Here is the idea that ties surfaces to everything still ahead. Any interface carries an energy per unit area, so a system can lower its total energy by shrinking its total interface area — and that single fact is a genuine driving force. It is why a foam of small grains coarsens into fewer, larger ones (grain growth), why loose powder sinters into a dense solid, and why a liquid drop beads up or spreads to balance the competing surface energies of solid, liquid, and vapour (wetting, set by the contact angle in Young's equation). The free surface is just the extreme case: an interface whose far side is empty vacuum, and so the highest-energy boundary of all.
The rest of this rung turns inward, to the boundaries where crystal meets crystal rather than crystal meets vacuum. Next comes the grain boundary, the mismatched seam between two grains of the same phase but different orientation — like floor tiles laid at different angles — split into gentle low-angle walls of dislocations and abrupt high-angle boundaries. Then the coincidence-site lattice explains why a few special misorientations are unusually low in energy; the twin boundary is a mirror-related, nearly perfect, very low-energy seam; the stacking fault and antiphase boundary you already glimpsed among the dislocations return; and finally the interphase boundary between two DIFFERENT phases — coherent, semicoherent, or incoherent — opens the world of epitaxy and lattice misfit. Every one is a 2D defect with its own energy, and because its far side is still crystal, each is gentler than the raw free surface we started with.