Boron Doping
Boron doping is the deliberate introduction of boron atoms into silicon to make it p-type, in which current is carried mainly by positively charged holes. Boron has three valence electrons, one fewer than silicon's four, so when it replaces a silicon atom in the crystal lattice it leaves a missing bond that readily accepts an electron from a neighbour; boron is thus an acceptor, and each atom contributes one mobile hole at room temperature. Boron is by far the dominant p-type dopant in silicon technology, used from the substrate wafer itself to the finest transistor regions.
Boron reached the factory floor through diffusion furnaces and a handful of source chemicals. In the diffusion era of the 1950s and 1960s, boron was delivered from boron trioxide, a solid or deposited glass that releases boron at furnace temperature, or from boron tribromide, a volatile liquid carried into the tube by a stream of gas, which reacts with oxygen to form a boron-rich glass on the wafer surface. Diborane, a toxic and flammable gas, and boron trichloride serve as gas sources in deposition and etching processes. In every case the wafer is held at roughly 900 to 1100 degrees Celsius so that boron from the surface glass diffuses into the silicon, as described under solid-state diffusion. The attraction of boron is that it is small and moves at a convenient rate, fast enough to reach useful depths in an hour yet slow enough to be controlled.
The ceiling on how much boron can be packed in is set by solid solubility, which is on the order of 10 to the 20 per cubic centimetre. Above this level, extra boron no longer sits on lattice sites as an active acceptor but clusters or precipitates, so the electrical activity stops climbing even though more atoms are present. Roughly 10 to the 20 per cubic centimetre is already a few tenths of one percent of the atoms in silicon, enough to make the material behave almost like a metal, and the highest practical levels are used under contacts and for heavily doped regions where low resistance matters. At the other extreme, lightly doped wafers with only about 10 to the 15 per cubic centimetre or so of boron are the starting substrates for many integrated circuits.
Boron's role in devices depends on which transistor is being built. In a bipolar transistor of the n-p-n type, the thin base is a boron-doped layer sandwiched between n-type emitter and collector, and the thinness and doping of that base set the current gain. In a metal-oxide-semiconductor circuit, boron forms p-type wells that hold n-channel transistors, the source and drain regions of p-channel transistors, and the channel-adjusting implants that set threshold voltage. It is also the standard dopant in substrate wafers, added to the molten silicon when the crystal is pulled so that the entire wafer is uniformly p-type, a practice dating to the earliest crystal-growing work in which dopants were introduced into the melt.
Ion implantation largely replaced diffusion as the way of placing boron, bringing its own subtleties. Implanted boron ions, accelerated to energies from a few hundred electron volts to hundreds of kilovolts, land at a chosen depth, but they are light and can travel along open channels between rows of atoms in the crystal, an effect called channeling, and the damage they cause must later be annealed away. During the anneal, boron can diffuse unexpectedly fast because implant-generated defects enhance its motion, a transient effect that engineers must contain as junctions have become only tens of nanometres deep. Heavier boron difluoride molecules are sometimes implanted to keep the boron shallow.
Several pitfalls follow from boron's chemistry. Boron tends to segregate into growing silicon dioxide, which depletes the silicon surface of boron during oxidation and changes the surface doping, a point that device designers must compensate for. Boron-doped glasses are used as insulators, but the boron-10 isotope absorbs thermal neutrons and has been linked to soft errors in memory, which led manufacturers to favour boron depleted in that isotope. And in solar cells, boron-doped Czochralski silicon is known to lose efficiency after illumination through a boron-oxygen defect, one reason that p-type wafers using gallium instead, or n-type wafers, have gained interest. A frequent misconception is that doping adds a new material: at parts per thousand or less, boron only perturbs the silicon, leaving its crystal structure intact.
Read boron doping as the standard way to make silicon p-type, a small admixture of missing electrons that supplies the holes carrying current and so define the wells, bases and junctions of nearly every chip.