Copyright Mark P D Burgess 2008
A taxonomy of transistor types begins with the junction type:
1. Metal-Semiconductor transistors from 1947
2. P-N Semiconductor junction transistors from 1950
The first transistor, the point contact transistor developed at Bell Laboratories and the later surface barrier transistor developed by Philco are examples of metal-semiconductor types. From an evolutionary point of view these were dead ends. They were rapidly relegated by the junction transistor invented by William Shockely.
The early junction transistors were variants of the grown junction types in which single crystals of germanium (originally) and later silicon were pulled from a melt and manipulated to create P-N junctions. These variants were
1. Double Doping
2. Rate Growing
4. Grown diffused
Since this approach created large single crystals, transistors had to be cut from the crystal in the form of a small bar that had the base layer at its centre. The early grown junction transistors had very poor frequency response due to the width of the base and base-collector capacitance. Rate growing and the later variants gave significantly improved performance but were extraordinarily expensive to produce and only viable for military use.
An alternative approach started with a wafer of P or N-type semiconductor and diffusing in dopants in a controlled manner to create an NPN or PNP transistor. These approaches were
1. Alloy junction transistor
2. Diffusion transistor
3. Drift transistor
4. Planar technologies
The alloy junction transistor was better suited to mass production and became the method of choice early on. But it suffered from poor performance in the RF range. This was solved by moving from alloy doping to gas phase (diffusion) doping which enabled better control of transistor geometry. Diffusion enabled the drift transistor: where the doping profile across the base could be graduated to improve the rate of carrier migration and hence frequency response.
Planar technologies invented at Bell (the mesa transistor) and commercialised by Fairchild (planar transistor) provided the foundation for the modern integrated circuit era. Photolithography enabled the fabrication of many transistors on a single wafer driving huge cost reductions.
The first point contact transistor was developed from high purity polycrystalline germanium but all commercial junction transistors were based on very high purity single crystal semiconductor. The necessary technologies were:
1. Chemical production of high purity semiconductor
2. Single crystal production
3. Zone refining for improved purities
The grown junction types discussed here are double doping, rate growing, meltback (and meltback diffused) and grown diffused. The double doping approach was crude and gave way to the later improvements. Its successors all had the advantage of accelerating base fields and easier control over base width for improved high frequency performance.
The story of the first grown junction transistors is told in the companion article on this site Semiconductor History: Faraday to Shockely in the context of the overall conception of the Junction transistor by Shockley from late 1947 to early 1948.
Following the materials choice for the point contact transistor, Shockley wanted a junction transistor made from polycrystalline germanium. But Gordon Teal, a Bell Labs physical chemist was keen to extend work on single crystal germanium proposing that minority carriers would have longer lifetimes in single crystals. In pursuing this approach he and Morgan Sparks developed a method for pulling single crystals from an N germanium melt creating the P base layer by “pill dropping.” In this example a P-type germanium “pill” is added to the melt converting it to P-type overall. The crystal is grown for another 0.005 inch and then converted back to N-type creating a large scale NPN structure. Morgan Sparks records the method in his lab book: “Yesterday we pulled an N-P-N junction by doping twice (once with Ga-Ge second with Sb-Ge during a pull of a crystal. The unit was cut longitudinally…by etching away the N sides we were able to solder to the P-type bridge.”
Shockely writes from 1976 “This non-photogenic device did perform according to the theory but had a wide base and poor frequency response and provoked little interest. For about nine months, the efforts to improve junction transistors were practically negligible at the laboratories.”
However, by mid 1951 a range of viable transistors had been made (Shockely 1951). These operated with power levels from milliwatts to 2 watts. Their collector junction capacitance limited their operating frequency to the audio range.
Controlling the base layer in the Grown Junction process is difficult requiring careful orchestration of the pulling rate, temperature, relative volumes of the melt and the pills, pill dopant concentration and the timing of the addition. There is plenty to go wrong and the double doping method became obsolete.
The more useful variants of the Grown Junction transistor (Rate Growing, Meltback and Grown Diffused) depend on kinetic dopant segregation effects that exist between a doped melt and the crystal in formation. These effects are:
(1) Impurities and dopants are strongly partitioned to the melt (following the classical principle in chemistry that crystal formation always results in significant purification: ie impurities are left in the liquid phase).
(2) The extent of incorporation of impurities and dopants depends on the rate of crystallisation. Fast crystallisation rates lead to incorporation of more impurities; Slow crystallisation results in purer crystals. More importantly for the formation of P-N junctions different impurities respond differently to crystallisation rates.
Hall 1952 reported that the manipulation of doping levels and crystal growth rates could lead to the formation of N or P germanium from a melt that contained both impurities. The figure shows the segregation constants for aluminium and gallium. (The segregation constant is the ratio of the element in the solid phase to the liquid phase.)
“Most P-Type impurities such as Ga and In show much smaller changes in segregation constant [with growth rate]. It follows that an ingot grown from a melt doped with the proper amount of antimony and either gallium or indium will be P-Type when grown slowly and N-Type when grown quickly. By cycling the growth rate it is possible to produce a large number of uniform and evenly spaced P-N junctions, each cycle producing two junctions having somewhat different chacteristics.” This is because the segregation constant of gallium is about 0.1 at low growth rates whereas the segregation constant for aluminium is 0.003: thus 30 times more gallium is incorporated on crystallisation at low growth rates (for equal concentrations of both).
The Rate Grown method eliminated the pills: both donor and acceptor dopants are present in the melt. Their relative incorporation into the growing crystal depends on crystallisation rate which in turn can be controlled by the amount of heat applied to the melt. Cooling accelerates crystallisation and excess heat causes re-melting. When there is no crystallisation a narrow P-type zone is established. By repeating such a process a succession of NPN structures can be grown into the crystal. These are then recovered by dicing the crystal into small bars that can be assembled into individual transistors.
The Meltback process is based on the relative kinetics of incorporation of P or N-type dopant according to crystallisation rate.
In this case the a single crystal made with both dopants (but overall is N-type) is diced into bars that will become an individual transistor. One end of the bar is heated to its melting point. Surface tension keeps the molten end attached. The heat is removed and a P base layer is formed first followed by an N-type zone as the molten drop re-crystallises. This process yields NPN transistors.
A variant that gives improved high frequency performance is the meltback-diffused process. Lesk 1957 gives an example of a PNP transistor formed from a bar containing two P dopants and one N-Type. After meltback and crystalisation a PNP structure is formed in which one of the three dopants predominates in each region.
Both meltback grown junctions feature accelerating base fields. Lesk’s experimental transistors showed a gain of 15db at 60mhz.
In the Grown Diffused method the initial melt doping is controlled for the desired collector conductivity. Base and emitter dopants are added simultaneously but their incorporation is dictated by their relative diffusion rate to the crystallising interface.
The alloy junction transistor was invented by John Saby at General Electric and similar developments were undertaken by Jacques Pankove at RCA. Inventorship had to be established in the US Courts as RCA filed on Pankove’s work one day ahead of General Electric in June 1952.
They were initially PNP types and commenced with a wafer of N-type germanium, typically doped with antimony which became the base. Some details of production are set out in Pankove’s patent [Pankove 1952].
The transistor base is made from doped mono-crystalline germanium typically of 2-5 ohm-cm and of N-Type for PNP transistors. The crystal is diced into wafers and then acid etched to remove surface defects caused by dicing. The etched wafer are 3-6 mils in thickness. Indium dots are prepared 10 mils thick and 15 mil for the emitter and 80 mil for the collector (mil = 1/1000 inch). The collector dot is applied first and adhered by firing at 250C for one minute The emitter dot is applied and the wafer fired at 400-500C for 10-20 minutes.
The firing process is enough to melt the indium which alloys with the germanium. The alloy has a lower melting point than pure germanium and two molten puddles of alloy form under the indium dots. On cooling the germanium alloy re-crystalises and re-forms a single crystal. As indium is an acceptor P-type regions form under the indium dots creating the PNP structure. By managing the wafer thickness and the alloying time the width of the base region can be controlled.
Gottlieb E Gutzwiller F Jones D Lowry H Snyder G Stasior R Sylvan T 1959 General Electric Transistor Manual Fourth Edition
Hall R 1952 P-N Junctions Produced by Growth Rate Variation Phys Rev 88 139
Lesk I Gonzalez R 1957 High-frequency transistors by the diffused-meltback process employing three impurities Electron Devices Meeting, 1957 International 3 117
Pankove J 1952 US Patent 3005132 Transistors Filed 13th June 1952 Patented 17th October 1961
Shockley W Sparks M Teal G 1951 P-N Junction Transistors Phys Rev 83 151-62
Shockely W 1976 The Path to the Conception of the Junction Transistor IEEE Trans on Electron Devices ED-23 July 597-620
Picture Credits General Electric Transistor Manual 4th Ed