Diffused-Base Transistor

A diffused-base transistor is a junction transistor whose base layer is formed by diffusing dopant atoms into the surface of a semiconductor wafer at high temperature, rather than by growing the layer in a crystal or alloying metal into it. The wafer itself serves as the collector, the diffused skin becomes the base, and the emitter is added on top by a second step. Developed at Bell Labs in 1954 and 1955, with Calvin Fuller and Morris Tanenbaum among the key names, the method gave engineers a base width they could control to a fraction of a micrometre, and that single capability turned the transistor from a laboratory curiosity of kilohertz and low megahertz into a device for hundreds of megahertz.

Diffusion replaced alloying because alloying could not make the base thin or repeatable enough. In an alloy-junction transistor, pellets of indium or a similar metal are melted into opposite faces of a thin germanium slab, and the base width is whatever remains between the two alloyed regions, a few thousandths of an inch for most production devices and hard to hold to a tolerance that mattered. Diffusion works in the opposite direction. A wafer is heated in a furnace in the presence of a vapour containing the desired dopant, atoms move into the crystal by random thermal hopping, and the depth they reach grows roughly as the square root of the diffusion coefficient times the time. Both temperature and time can be set precisely, so junction depths of a micrometre or less became a matter of furnace discipline rather than craft.

The diffused base brought a second benefit that nobody had to design: it was graded. Because the concentration of diffused dopant is highest at the surface and falls with depth, the base has a doping gradient and therefore a built-in electric field that speeds carriers across it, the effect that gave the drift transistor its name. A diffused-base device thus combined a narrow base with a field-assisted transit, which is why the frequency improvements were so large. Fuller's earlier work with diffusion at Bell Labs, for example in making p-n junctions for the silicon solar cell announced on 25 April 1954 with Chapin and Pearson, had already shown how well diffusion could form shallow, high-quality junctions on a large scale.

Tanenbaum's name is attached to the silicon version of the story, and the distinction between versions matters. Morris Tanenbaum at Bell Labs fabricated an early silicon transistor in January 1954, which was a grown-junction laboratory device rather than a commercial product like the one Texas Instruments announced on 14 May 1954, and he then applied diffusion to silicon and germanium to make high-frequency devices. In many first diffused-base designs the base was diffused but the emitter was still an alloyed dot of metal, so the device was a hybrid, a distinction that a double-diffused transistor then closed by making the emitter diffused too. A common mistake is to treat diffusion as a single invention by a single person; it was a sequence of improvements by teams who learned how to make deep and shallow junctions on demand.

The method had limits that shaped what came next. The geometry was still crude by later standards: the diffusion covered the whole wafer surface, so the base layer extended across the entire chip and had to be cut or etched away to define individual devices, which led directly to the mesa transistor and to Hoerni's planar approach, in which windows in an oxide layer define where the dopant enters. Germanium also allowed only modest temperatures and was a poor host for diffusion masks, which is one reason the industry's attention moved to silicon, where an oxide layer could block diffusion after the Frosch and Derick discovery at Bell Labs in 1957. Diffusion also introduces impurity profiles that are hard to alter after the fact, so the process has to be right the first time.

The diffused base is the direct ancestor of how bipolar transistors, and later integrated circuits, are built. Diffusion by furnace, followed by ion implantation, remains the standard way of placing a dopant at a known depth, and the idea that process parameters, not hand craft, define a device is a part of the planar integrated circuit manufacturing philosophy. It also introduced the vocabulary designers still use: sheet resistance, junction depth, surface concentration and diffusion length. A modern chip's source and drain regions are formed by methods that evolved from these first furnace steps, even though the transistor type is now usually a MOSFET rather than a junction device.

Read diffused-base transistor as the moment the base stopped being a thickness left over from alloying and became a depth deliberately set in a furnace.