Double Diffused Silicon Mesa 1956 Drift Transistor

# The Double-Diffused Silicon Mesa Transistor: A Graded Base That Lets an Electric Field Do What Diffusion Alone Could Not

In 1956, Bell Labs and RCA moved the double-diffusion technique this project has already documented on germanium onto silicon, and paired it with a change that was not just a material swap: deliberately grading the base doping profile so a built-in electric field, not diffusion alone, carries injected carriers across the base. The 1954 diffused-base germanium process set base width as the difference between two independently controlled Gaussian depths, and that base region was doped as uniformly as the predeposition-and-drive-in physics naturally produced. The 1956 drift transistor keeps the double-diffusion and mesa-isolation machinery essentially intact, but exploits the gradient already present in a drive-in profile on purpose — turning a side effect of the diffusion process into the thing that makes the finished device faster.

The physics is a direct consequence of how a drive-in anneal already works, pushed deliberately further. A Gaussian drive-in profile is never perfectly flat; concentration naturally falls with depth, and that fall implies a built-in electric field even in an otherwise ordinary diffused base, because a doping gradient at thermal equilibrium always produces one. The 1954 process treated that field as incidental, small enough relative to a moderate base width that diffusion still dominated carrier transport. This process instead chooses drive-in conditions specifically to make that gradient steep, so the resulting field becomes strong enough to sweep injected carriers across the base by drift — a transport mechanism whose transit time scales with base width to the first power, not the second, which is the direct reason a drift-graded base can be meaningfully faster than a uniformly diffused one of the same physical thickness.

Silicon is not a cosmetic substitution for germanium in this process; it is the material the drift architecture actually needed to matter. Germanium's narrower bandgap lets leakage current grow fast enough with temperature that devices built from it become unreliable well below 100°C, a ceiling that limits how aggressively any germanium device, drift-graded or not, can be pushed. Silicon's wider bandgap keeps leakage low to a much higher operating temperature, which is the condition under which a drift transistor's speed advantage is actually worth having — a device fast enough to matter but still thermally limited the way germanium devices were would have traded one bottleneck for another rather than removing one.

This device does not introduce a new fabrication philosophy; it is the moment this project's own double-diffusion lineage and the material that would actually dominate the industry finally arrive in the same part. Every earlier step this project has documented — the point-contact's spring-loaded whiskers, the grown-junction's pull-timed dopant switch, the alloy-junction's fired disks, the surface-barrier's jet-etched membrane, the diffused-base's two independently scheduled furnace anneals — was germanium work, because germanium was what the field had learned to purify and dope reliably first. Silicon's own diffusion chemistry simply needed the double-diffusion technique this project has already traced in full before a drift-graded base on silicon was within reach, and the year that technique crossed onto silicon happened to be the same year the field's three founders were recognized for the work that started all of it, and the same year the company that would train most of the engineers who built everything after it opened its doors.