Oxide Masking
Oxide masking is the use of a layer of thermally grown silicon dioxide on a silicon wafer as a barrier that stops dopant atoms from entering the silicon wherever the oxide is present, so that doping happens only through windows cut in the oxide. It is a deceptively simple idea: grow an oxide skin, cut holes in it where junctions are wanted, and put the wafer in a diffusion furnace, where dopants reach the crystal only through the holes. Because the oxide is glass-like and survives the furnace, it is a permanent-until-removed mask, and it became the geometric foundation of modern silicon device fabrication.
The discovery was made at Bell Labs in 1957 by Carl Frosch and Lincoln Derick, and it was an accident. They were studying the diffusion of dopants into silicon wafers in a furnace, and an oxide layer formed on the surface, in the account usually given, unintentionally. When they examined the wafers, they found that the oxide-covered regions had not been doped as expected, and that the oxide, rather than being a nuisance, could protect the silicon underneath. Their paper in the Journal of the Electrochemical Society later in 1957 described surface protection and selective masking during diffusion in silicon. What made it important was not the observation alone but its meaning: a silicon wafer could carry many closely spaced junctions defined by a pattern, rather than one junction across the entire surface.
The planar process of 1959 was built on this foundation. Jean Hoerni at Fairchild Semiconductor recognised that if oxide was kept in place, the junction edges emerging at the surface would be buried under it and protected from contamination, instead of being left bare as in mesa devices. Photolithography, adapted for transistors by Jay Lathrop and James Nall at the US Army's Diamond Ordnance Fuze Laboratories in 1957 and 1958, supplied the means of opening precise windows. Together, oxide growth, window opening and diffusion could be repeated: one window for the base, another within it for the emitter, with every junction defined by a mask rather than by hand. That repeatable, pattern-defined sequence is what allowed many transistors and, soon afterwards, whole circuits to be made on a single wafer.
Masking and passivation are related but distinct benefits of the same oxide. Mohamed Atalla at Bell Labs and his colleagues showed in 1957 to 1959 that a thermally grown oxide dramatically reduced the surface states that had plagued silicon devices, and their work on stabilising silicon surfaces led to the MOS transistor demonstrated by Atalla and Dawon Kahng. Masking is about stopping dopants during processing; passivation is about electrical stability afterwards. The same oxide, a few tenths of a micrometre to about a micrometre thick, could do both jobs, which is part of why silicon, with its excellent native oxide, displaced germanium, whose oxides are soluble and poor.
The mask is good but not perfect, and the pitfalls are specific. Silicon dioxide blocks boron, phosphorus, arsenic and antimony well, but gallium and aluminium diffuse rapidly through it, so they are poor choices for oxide-masked processing. Dopants do penetrate the oxide eventually, so the oxide must be thick enough for the time and temperature of the diffusion, and boron tends to be absorbed into the oxide during growth. Windows also diffuse sideways, so a junction extends a little beyond the opening. A common misconception is that the photoresist is the diffusion mask; resist cannot survive furnace temperatures, which is why oxide serves as the hard mask, with resist only defining the oxide. Today, oxide and nitride hard masks, local oxidation for isolation, and implant masks all descend from the same idea, even though implant doping uses resist and thinner films.
Read oxide masking as the accidental discovery that a wafer's own glass skin can be carved into stencils, turning doping from a blanket treatment into a pattern and so launching the planar process.