Making a 200 Series NPN Grown Junction Transistor
For the full history of Texas Instruments on this site please return to The Early History of Texas Instruments Semiconductors.
Grown junction transistors were made on crystal pullers which were used to pull a solid single crystal of germanium from a crucible of molten germanium. The NPN structure was achieved by starting with N type germanium, changing it to P-Type then back to N-Type while pulling the crystal.
Ed Millis recalls details of the crystal pullers:
“The crystal pullers were ingenious phone booth-sized machines designed by Boyd Cornelison, TI’s sagacious semiconductor chief scientist. The crystal puller heated a teacup-sized crucible of germanium to red heat, or about 950 Celsius, and then dipped a small piece of special germanium called a “seed crystal” into the molten pool. As the crystal was spun slowly and simultaneously drawn slowly out of the melt, a single crystal ingot would “grow” on its end. And that’s the stuff you needed to make transistors. Just to make it dangerous, the melting and pulling had to be done in a hydrogen atmosphere, and keeping air out of it was necessary to keep it from exploding.” [Millis 2000]
The width of the crystal was controlled by the melt temperature and the rate the crystal was withdrawn from the melt. Typically seed crystals were 125 mils diameter and the crystals grown to a one inch diameter.
The first portion pulled would typically become the collector, having the lowest conductivity.
The P-type base layer was created by adding a pellet of P-Type impurity such as phosphorous or gallium to the melt. The period of time the melt remained P-Type and the pulling speed determined the width of the base.
Then the melt was made N-Type by adding a pellet containing arsenic or antimony. This portion had higher conductivity and became the emitter.
In calculating the final result engineers had to allow for the manner in which impurities preferentially remained in the melt. See, for example, a discussion on segregation constants and how diffusion phenomena were also used to make the vastly improved grown diffused types (developed by TI) and other variants such as rate grown and meltback types.
Next the crystal needed to be sliced into thin bars that would become the transistor. Bill Bower recalls: “The next step was to cut off the upper and lower N regions to yield a round NPN wafer about ¼ inch thick, with the P base in the center of the sandwich. Using a multi-bladed diamond saw, the wafer was then sliced into narrow rectangular NPN bars. These bars were chemically etched to remove surface damage.” [Ward 2001]
Lastly each bar was mounted on a header and canned as described by Ed Millis: “When I began in 1954, the NPN transistors were made by mounting an NPN bar, maybe 15 or 20 mils square and 250 mils long in a formed header, by soldering the ends of the bar (the collector and the emitter) to the appropriate header posts by hand.
Then, in a gold bonding machine, a 1 or 2 mil gold wire, held in a pair of manipulator tweezers, was brought down to about the center of the bar, and raked back and forth while watching an oscilloscope pattern. The P junction was not visitble to the eye, so you had to feel for it, electrically. When you found it, you put the wire right in the middle of it and pushed the bonding button. It delivered just the right jolt of electriciity to melt the end of the wire into the P junction, and also into the adjoining N junctions on either side The wire from the junction had to be bent down and tweezer-welded to the base connection of the header.
The area around the wire bond had to be etched back to clean it up, and "Big Orange" was used. This was hydrofluoric acid, nitric acid, and bromine, as I recall. The color of it, from the bromine, resembled a Nehi orange soda. It was highly corrosive and the header metal parts and the ends of the bar where it was soldered in were masked with a red lacquer. This was yet another manual task done with a small artist's brush. The device was held, by holding onto the three leads with your rubber-gloved fingers, into a stream of etchant, and let it flow over the wire bond spot on the bar for a prescribed number of seconds. Not terribly exact, but it cleaned up the junction. Then the units were boiled in alcohol to strip the masking resist. This was very labour intensive, and led to the ladder process.
Then the device canned.”
Because the base width was very narrow it was likely that the bonded wire would overlap and short the base to the emitter or collector. To avoid this the wire was made P-Type. This meant it would be isolated from the emitter or collector by a PN junction and make an ohmic connection to the base. [Millis 2011]
References
Millis E 2000 TI, the Transistor and me Ed Millis Books, Dallas
Millis E 2011 Personal communications May 2011
Ward 2001 Bill Brower The Art and Science of Building the First Commercial Silicon Grown Junction Transistors
http://www.ck722museum.com/history/Transistormuseum/LectureHall/Brower/Brower_Index.htm