Single Crystal
A single crystal is a solid in which the atomic lattice runs unbroken from one edge to the other, with every atom sitting where the repeating pattern says it should. Silicon in a wafer is a single crystal of the diamond-cubic structure, with a lattice constant of about 0.543 nanometers, each atom bonded to four neighbors, and the pattern continuing without a break across a disk 300 millimeters wide. The contrast is with polycrystalline material, which is a patchwork of many small crystal grains in different orientations, and amorphous material, which has local bonding but no long-range order at all.
Grain boundaries are the reason single crystals matter electrically. Where two grains meet, the rows of atoms do not line up, leaving strained and broken bonds, essentially a sheet of dangling bonds and defects. Those sites scatter moving carriers, so mobility drops; they trap carriers and make recombination centers, so lifetime drops; and they act as fast paths along which dopants and metal atoms diffuse, so junctions become leaky and non-uniform. A transistor channel only a few nanometers across and a few tens of nanometers long, if it straddled a grain boundary, would have a threshold voltage and current wildly different from its neighbor, so integrated circuits need a substrate whose lattice is continuous under every device.
Single crystals are made by seeding: all of the material freezes onto one starting crystal. The Czochralski process pulls a boule from the melt on a seed, while the float-zone method carries a seed through a molten zone, and both reproduce the seed's orientation exactly. Gordon Teal and John Little pulled single-crystal germanium at Bell Labs in 1950, and the transistors of the 1950s showed how much cleaner the physics became once polycrystalline samples were left behind. The crystal's orientation is then chosen and marked on the wafer: most logic is built on the (100) plane, which gives a low density of interface traps and well-behaved oxides, while (111) wafers appeared in older bipolar processes.
Polycrystalline and amorphous silicon are not failures but cheaper tools for different jobs. Polycrystalline silicon, deposited as a film, forms the gate electrode in decades of MOS transistors, where only its conductivity matters, not carrier mobility. Multicrystalline silicon, cast in blocks, was long the cheap workhorse of solar panels, and hydrogenated amorphous silicon, deposited in thin layers on glass at low temperature, serves in thin-film transistors for display backplanes and in some thin-film solar cells. Their grain boundaries or disorder lower carrier mobility and lifetime, and so, in solar cells, efficiency, which is why the industry has largely moved to single-crystal wafers with passivated surfaces for the highest-efficiency modules.
Special wafers keep the single crystal where it is needed while changing what is underneath. Silicon-on-insulator wafers carry a thin single-crystal silicon layer over a buried oxide, which reduces parasitic capacitance and leakage, and are made with processes such as Smart Cut, which transfers a thin crystal layer by implanting hydrogen and splitting it off. Silicon on sapphire, grown epitaxially on a single-crystal aluminum oxide substrate, served radio-frequency and radiation-tolerant circuits. A misconception is that a polished wafer is flawless: real single crystals contain point defects, and may include dislocations, oxygen precipitates and stacking faults, and fabs inspect and engineer them. The aim is to keep them away from active regions, not to reach mathematical perfection.
Read single crystal as continuity of the lattice: when every atom has its place, electrons move predictably, junctions form where designed, and a billion nearly identical transistors can share a wafer, whereas grain boundaries and disorder, tolerable in solar panels, display backplanes and gates, would turn the same device physics into noise.