Surface States

Surface states are electronic energy levels that exist only at the surface of a crystal, where the perfectly repeating lattice of the bulk suddenly ends. Inside a silicon crystal every atom shares four bonds with its neighbors; at the surface, atoms are left with unpaired orbitals called dangling bonds, and these can capture or release electrons at energies that fall inside the forbidden band gap. Because the gap is exactly where a clean semiconductor has no allowed states, even a modest density of surface states can hold a large amount of electric charge, and that charge can control how the whole surface behaves electrically.

The theory of surface states came before anyone could build a device that needed it. Igor Tamm predicted localized surface levels in 1932 by solving the Schrödinger equation for a crystal with a sharp termination, and William Shockley found a second kind in 1939 using a different, more band-like model of the lattice. In the literature the two are still called Tamm states and Shockley states: Tamm states are tied to strongly perturbed, atom-like surface orbitals, while Shockley states arise when the bulk bands bend and cross near the surface. Real surfaces, with dangling bonds, reconstruction, adsorbed atoms, and oxide, usually mix both ideas, so the modern phrase surface states covers any electron level localized within a few atomic layers of the boundary.

Surface states explain one of the most famous failures in transistor history: the field effect that should have worked in 1947 and did not. Shockley hoped that a voltage on a metal plate held near a thin semiconductor film would pull in enough mobile charge to change its conductivity, a field-effect amplifier of the kind Lilienfeld had patented in the 1920s. The measured effect was far smaller than the calculation predicted. John Bardeen's 1947 surface-state theory explained why: electrons from the interior got trapped in surface states and formed a sheet of charge that screened the interior from the applied field, like a metal shield. That insight redirected the Bell Labs group toward injecting carriers through a contact, and within months it led Bardeen and Walter Brattain to the point-contact transistor of December 1947, a result that was later recognized in the 1956 Nobel Prize in Physics.

The solution came a decade later from an unexpected material property: silicon forms a remarkably good oxide. Carl Frosch and Lincoln Derick at Bell Labs noticed in the mid-1950s that a layer of silicon dioxide grown in a hot oxidizing furnace could mask the silicon beneath it from dopant diffusion. Mohamed Atalla then studied the interface itself between 1957 and 1959 and showed that thermally grown oxide bonds with surface silicon atoms, satisfies most of the dangling bonds, and leaves the surface electrically quiet; his group at Bell Labs published the stabilization of silicon surfaces by thermal oxide in 1959. With Dawon Kahng he then built the first working metal-oxide-semiconductor field-effect transistor, finally delivering the field effect that surface states had blocked in 1947. The same oxide underpinned the planar process, which is why this single interface sits at the root of both the MOSFET and the modern integrated circuit.

Interface traps did not vanish; they were merely tamed, and they remain a central quantity in device engineering. A good thermal Si-SiO2 interface still carries on the order of 10^10 trap states per square centimeter per electron-volt, compared with a density of surface atoms near 10^15 per square centimeter, so only a small fraction of the original bonds is left unsatisfied. Engineers anneal wafers in hydrogen or forming gas so that hydrogen caps those last dangling bonds, and they watch interface traps because the leftover states cause threshold-voltage shifts, 1/f noise, and the slow degradation known as bias-temperature instability. In today's transistors with hafnium-based high-k gate dielectrics, a thin silicon oxide layer is deliberately kept at the interface precisely to preserve that low trap density.

A common misconception is that surface states are simply contamination that cleaner processing would eliminate. Even in ultra-high vacuum, a perfectly clean silicon surface has dangling bonds and rearranges itself into reconstructed patterns such as the 7×7 structure of the Si(111) surface; the states are a consequence of the termination itself, not dirt. The practical question is therefore not whether surface states exist but whether they can be passivated, and silicon is unusually lucky in this respect. Germanium oxide is water-soluble and unstable, and many compound semiconductors such as gallium arsenide suffer Fermi-level pinning from surface states that no native oxide has been able to fix, which is one reason silicon, not germanium or gallium arsenide, dominates integrated circuits. The same logic returns in solar cells, where surface passivation layers raise efficiency, and in nanowires and 2D materials, where nearly every atom is a surface atom.

Read surface states as the price of ending a crystal: broken bonds put charge-trapping levels in the band gap, they once blocked the field effect and sent Bell Labs down the path to the point-contact transistor, and the thermally grown oxide that tamed them is the quiet reason the MOSFET and the silicon chip exist at all.