From Germanium to Silicon Chapter 2 A History …_1997.pdfChapter2 From Germanium to Silicon A...

Chapter 2From Germanium to Silicon

A History of Changein the Technology

of the Semiconductors

by Philip Seidenberg

Silicon has been the dominant semiconductor materialsince the middle 1960s. Today, probably 95% of allsemiconductors are fabricated in silicon, yet the first

transistor was a germanium device. Until 1960 most designengineers preferred germanium to silicon for computer logiccircuits, when, suddenly, germanium was out, and siliconwas in. What caused this abrupt shift to silicon? An answerto this question requires some understanding of how andwhy solid-state scientists went about their research.

This chapter explores the technical choicesconcerning the use of germanium and silicon assemiconductor materials made by scientists and engineersduring the period from 1947, the year the transistor wasinvented, until 1960 when the design shift from germaniumto silicon occurred. During this period, the scientists andengineers at Bell Telephone Laboratories (BTL) were theworld's leading investigators of the properties ofsemiconductors. With few exceptions, scientific research on

Seidenberg

the transistor was dominated by BTL; from 1947 through1959, twenty-four of the forty-four papers cited aspioneering papers in semiconductor device technology werewritten by BTL researchers.! The remaining twentyoriginated from fourteen entities, split evenly between thecorporate world and the academic/government complex. Infact, it was not until the end of the 1950s that other industriallaboratories begin to challenge the dominance of BTL insemiconductor research and development. This is in starkcontrast with the relatively minor role that BTL and itsmanufacturing arm, Western Electric, played insemiconductor manufacturing and sales. Through 1958Western Electric produced 1.315 million transistors,2representing less than 2% of the total amount of transistorsproduced by the semiconductor industry during this period.3

In spite of its secondary role as a semiconductorsupplier, the course of semiconductor research was shapedby the scientific and engineering achievements at BTL duringthe 1950s. While most of the semiconductor industryconcentrated on manufacturing germanium transistors anddiodes during this decade, BTL spent most of its researchdollars on silicon devices. This research, coupled with theresponse by the rest of the industry, led to the developmentof the surface-stabilized silicon transistor and diode, whichin turn signaled the replacement of the commerciallydominant germanium by silicon as the leading semiconductormaterial. In this chapter, I will argue that the precipitatingcause for silicon replacing germanium, particularly incomputer logic, was the use of a silicon dioxide film tostabilize the surface of silicon semiconductors. This was themost important factor in the development of the planarprocess, wh,ich led first to the dominance of the silicontransistor and then to the integrated circuit.

36

From Germanium to Silicon

There is very little historical research on theimportant change from germanium to silicon. Mosthistorians exploring the semiconductor field recognize thesignificance of the planar process, but treat lightly theprogress leading to this innovation. The seminal work onsemiconductor electronics refers very briefly to thestabilization of the silicon surface4 and another establishedhistorical work devotes slightly over half a page to the"surface stabilization of silicon by the oxide maskingprocess."5 An important book on technology transfer insemiconductors chronicles the changeover from germaniumto silicon but offers no insight on causes.6 None of thesetexts indicates that the replacement of germanium by siliconwas an important event in the history of technology, despitethe fact that many scholars consider the transistor as the mostpervasive invention of the last half of the twentieth century.Furthermore, the works that examine progress made in thephysics of semiconductor surfaces do not expound on itsinfluence on technology or industry. The volume onelectronics technology and the physical science in BTL's AHistory of Engineering and Science in the Bell Systemoffer a non-technical overview of what happened in surfacephysics and processing during the 1950s, yet BTL'shistorians present no explanation or interpretation of whatthe replacement of germanium by silicon meant tosemiconductor manufacturers and their computercustomers.7

When they do contemplate the demise of thegermanium semiconductor, historians often ascribe it to theinfluence of the Defense Department. It has been argued thataggregate funding figures for transistor development can beused as a gauge for military influence on semiconductor

37

Seidenberg

technology, and that military semiconductor technologyspilled over into the commercial economy.8 Although themilitary did fund silicon transistor development during the1950s, both military and commercial computermanufacturers almost exclusively used germanium diodelogic during this period because germanium diodes could bemass produced more reliably and less expensively than couldany other semiconductor. For example, the MinutemanMissile Program, which was probably the most importantmilitary program for semiconductor manufacturers in the late1950s, used germanium diode logic in its guidance system.Consider; also, that digital logic, the preferred choice amongcircuit designers for commercial computers, required largenumbers of inexpensive and reliable switching elements.The germanium gold-bonded diode fit their requirements.The tube and electromechanical relay were large powerhungry devices. The transistor, a three terminal device,suffered from low yields and unreliable performance whencompared to the two-terminal diode. Germanium diodeswere high-yield economical electronic switchesmanufactured in volume by reputable and reliable vendors.Clearly, the question of which material, germanium orsilicon, would dominate semiconductors was not driven bymilitary considerations.

Much of the literature about the military's earlyinfluence on the semiconductor field was written during the1980s, during the Reagan years of military spending. AsAlex Roland has observed, "A kind of presentism movesand infects much of this (1980s) literature." Aggregateresearch funding by the Defense Department favored silicondevices, but it was germanium that the military purchased forhigh-volume logic requirements. Even' at BTL, a virtual

38


hotbed of silicon research, practically all semiconductorsmanufactured through 1958 used germanium as the basic orstarting material.

Semiconductor development has been drivenprimarily by the study of the physics and chemistry of thebody and surfaces of its starting material. The body-theregion of the material away from the surface-is ahomogenous periodic structure. Because it poses lesserobstacles to theoretical comprehension than the surface, ithas been easier to analyze and measure. For a long timesolid-state theorists implicitly assumed that the properties ofsemiconductors were determined strictly by what happenedin the body of the material. The surface-construed here asthe boundary area between the body of the semiconductorand any other medium-does not have the symmetry andperiodicity of the body. Only a few atomic layers deep, thesurface influences profoundly the semiconductor action inthe body. These atomic layers of material foreign to thebody often are difficult to identify and neutralize.

Early efforts to understand semiconductor actiondepended largely on empirical inquiry. Theoretical conceptswere not formulated until the introduction of quantummechanics into semiconductors and the explanation for thebehavior of electrons in metals and semiconductors.AT&T' s corporate strategy for basic research at BTL gelledin the period 1925 to 1935 after the introduction of the newquantum physics. The idea of applying basic science totechnology became embedded in BTL strategy. As historianLillian Hoddeson has noted, "by the mid-1930's theproblems, approaches and atmospheres of fundamentalresearch at Bell Labs were remarkably similar to those inuniversity laboratories."9 Mervin Kelly set the research and

39

/

Seidenberg

development (R&D) parameters for BTL after World War II.Kelly, the vice-president of BTL, decided to focus R&Dactivities on semiconductors in order to develop a solid-stateamplifier for telephone applications. Before the war, Kellyhad been interested in substituting electronic relays formechanical relays in telephone exchanges."lO The SecondWorld War changed BTL's focus from peacetime research towartime engineering.

The war accelerated semiconductor development asscientists were transformed into engineers to work ongermanium and silicon crystal rectifiers for radar and radioequipment. Scientists and engineers selected germanium andsilicon for military research because they were elementalsemiconductors with an orderly atomic structure.Germanium and silicon crystallize in a diamond lattice,consisting entirely of one type of atom. They possess a lesscomplex structure than selenium or certain intermetalliccompounds such as copper oxide, which were usedextensively in rectifiers from the 1920s. Scientists at BTLanticipated easier chemical processing and fewer structuraldefects in materials that showed the least crystal complexity.

After World War II, BTL continued its pre-warresearch strategy. In the post-war social context, technologymeant progress. Radar had saved Britain from the Luftwaffein 1940. The atomic bomb had crushed Japan in 1945.Technology could win the peace as it had won the war, withthe United States lighting the way for the world. FrederickE. Terman, the Stanford University professor who headedtop-secret radar research at Harvard University during thewar, summed up the feeling of the scientific communitywhen he said World War II showed "science and technologyare more important to national defense than masses of men"

40


and "how essential the electron was to our type ofcivilization."11 It is in this environment that BTL embarkedupon a sustained research-and-development effort whichdrove progress in the semiconductor industry during the1950s.

Solid-state physicists interested in doing research inbasic science in the post-war period were faced with onlyone real choice if they preferred to work in industry ratherthan academia; BTL was the premier industrial laboratory forsolid-state research. The management at BTL was intent onattracting talented researchers and creating an atmosphere offellowship12 and the result was a research facility whichcould be compared favorably with those of the leadinguniversities in the United States. 13 George F. Dacey, whoworked for William Shockley, recalls that Kelly wasinsistent on recruiting only the best people that could befound. Shockley, who had joined BTL in 1936, already hadachieved a reputation in the field and served as a beacon toattract additional bright solid-state scientists. 14 MorganSparks, who took over Shockley's group in the mid-1950s,was impressed with Shockley's great insight as a theoreticalphysicist. Shockley was not too concerned with themathematics of a problem as long as he understood most ofthe figures. Shockley also evinced a great interest in deviceapplications 15 and appeared to approach a problem with aright mix of the theoretical with the practical. Dacey andSparks recall the frequent meetings, both formal andinformal, held at BTL to discuss the almost daily appearanceof new semiconductor phenomena ensuing from solid-stateresearch. Progress was so rapid in semiconductors that itdominated the meeting schedules. By 1950, Shockley hadabout twenty people in his solid-state physics group, which

41

Seidenberg

ultimately grew to nearly one hundred by 1960 under thedirectorship of Dacey, one of Shockley's successors.

Typical of the people drawn to BTL after 1950 wasC.G.B. Garrett. Garrett, who was a graduate of CambridgeUniversity, and an instructor in physics at HarvardUniversity, joined BTL in 1952 to do research onsemiconductor surfaces because he felt that it could result inprofessional recognition. Moreover, BTL could afford topay well and the laboratory atmosphere was collegial.Technology diffused easily and rapidly among all researchand development groups. Regularly scheduled, topical andad hoc meetings characterized the diffusion of semiconductorknowledge within BTL. It is in this ambiance thatsemiconductor research flourished. 16

A tabular view of the major milestones in transistorresearch and development between 1947-1959 reveals theimpact of BTL's research in this area (see Appendix 1). Ofthe ten major events instrumental in the demise ofgermanium as the preeminent starting material, sevenoriginated at BTL while the other three were derived fromBTL research activities. Of course, there was considerablymore happening in semiconductor research and developmentat BTL during this time besides these important events. Forexample, BTL researchers developed the phototransistor(J.N. Shive, 1950), the junction field-effect transistor (G.C.Dacey and I.M. Ross, 1952), the thermocompression wirebonder (O.L. Anderson, H. Christensen and P. Andreatch,1956), and other significant process and product innovationsnot directly connected to the history of the replacement ofgermanium by silicon.

One should not conclude these milestones intechnological innovation were part of a continuous stream of

42


BTL's successes, unaccompanied by individual failures andfrustrations. However, the rate of technological progress atBTL during this period did allow the semiconductor industryto grow to about one half billion dollars by 1960.17

Compare this to the inability of solid-state scientists andengineers to agree on common concepts for a basis ofresearch in the fifteen years from 1931, when the theory ofthe electronic semiconductor based upon quantum mechanicswas advanced. In a review of semiconductor research in a1955 issue of the Proceedings of the Institute of RadioEngineers, G.L. Pearson and Walter Brattain noted that "ittook about fifteen years for the full light to dawn" and thatthere were "many blind spots in the working concepts aboutsemiconductors in the nineteen thirties."18

The quality and quantity of the researchers at BTL,operating in an atmosphere of easy communication andunfettered disclosure, contributed considerably to thesemiconductor industry's high rate of growth during the1950s. The strategy of the management to encourage theresearchers to be independent promoted basic solid-stateresearch. Brattain remembers that after many years ofresearch trying "to understand what was really going on inthe simplest of semiconductors, silicon and germanium," hebegan "to lose faith." However, he felt "no pressure frommanagement to continue or to change fields.".

What had influenced Shockley and his twocolleagues, Brattain and Bardeen, to join BTL andconcentrate on the study of semiconductor surfaces? In1972, in an issue of the Bell Laboratories Recordcelebrating the 25th anniversary of the transistor, the threescientists recalled the personal interests that led to thediscovery of the transistor. The leader of the group, William

43

Seidenberg

Shockley, who obtained his doctorate by studying thebehavior of electrons in crystals, was attracted to BTL in1936 by the opportunity to work with C.J. Davisson, aneminent physicist in the field of electron behavior.19 WalterBrattain, the oldest of the three, evinced an early interest inquantum mechanics and entered the field of surface physicsat BTL in 1929 after receiving his doctorate. In 1937Brattain met Davisson when the latter used Brattain'slaboratory to demonstrate his work on electron behavior.John Bardeen, who joined BTL in 1945, had no experiencewith semiconductors, but prior to the war he had beeninterested in the theory of metals. He was anxious to "returnto solid-state physics after five years at the Naval OrdnanceLaboratory in Washington, D.C."20 Trained originally as anelectrical engineer, he had returned to school at Princetonafter several years in industry to obtain his doctorate inmathematical physics, drawn there by the opportunity tostudy quantum theory under Einstein.

All three scientists attributed the success of theirresearch to the supportive atmosphere created by the BTL inwhich management allowed those in research to spend manyyears, much money, and the talents of many people toexplore solid-state physics. From their recollections, itappears that all three were motivated by an interest in surfacephysics and electron behavior, and an opportunity to studyand work with eminent scientists in their chosen fields. Thephysical reality of the microscopic world of atoms openednew vistas for students of physics, particularly thoseexploring electron behavior in solids. On the 25thanniversary of the transistor, John Bardeen recalled that heand Walter Brattain discovered the transistor as a result offollow-up experiments on some consequences of Bardeen's

44


published investigations of semiconductor surfaceproperties. 21 Shockley, who with Bardeen and Brattainreceived the Nobel prize in 1956 for the invention of thetransistor, had theorized in 1939 that localized states canexist at the surface of a semiconductor.22 However, solidstate physicists did not realize the importance of thesesurface states at the time. The accepted theories ofsemiconductor action formulated prior to World War IIassumed that rectification was closely related to the polarityand magnitude of the contact potential at the interfacebetween a conductor (metal) and a semiconductor, orbetween two different s,emiconductors. The polarity andmagnitude of the contact potential depended upon thedifference in work functions of the metal and thesemiconductor. However, experimental measurements didnot confirm the predictions of the theory. There was littlecorrelation between the work function of the metal and thecontact potential.

This perplexing phenomenon was made whollyintelligible by Bardeen's hypothesis in 1947. Surfaceproperties rather than the contact potential at the interface ofthe metal and the semiconductor were the governing factorsfor device rectification. Bardeen theorized that electronsfrom the body of the semiconductor material become trappedand immobilized at the surface, repelling other electrons inthe conduction band. This produced a layer of depletedconductivity below the surface.23 He further theorized thatsurface states occur when the body of the semiconductormaterial is abruptly terminated, allowing electrons and holes(the absence of electrons) in the forbidden energy gapbetween the valence band and the higher-energy conductionband. Normally, in the body of the semiconductor material,

45

Seidenberg

electrons exist in the conduction band and holes in thevalence band. Between these two bands is the forbidden gapwhere neither electrons nor holes reside unless trapped byimpurities (surface states tend to act like impurities inpermitting electrons and holes to have energies in theforbidden gap). Trillions of surface states exist on a squarecentimeter of semiconductor material, dominating theelectrical properties of the structure. These surface stateswork against any electric fields or currents applied inaccordance with circuit design and also work to neutralizeany desired imbalance in electrons and holes by tending toreturn the semiconductor to its condition before theapplication of the electric field or currents. Realizing thatthese surface effects can dominate the body properties of thesemiconductor, Bardeen and Brattain conducted experimentsto change the surface potential of germanium. It was duringthis research that they invented the transistor24 (Appendix2a). Thus, the mutual interest of these two physicists inelectron behavior in solid states led them to this importantdiscovery.

The invention of the transistor sparked an heightenedinterest in semiconductors. Prior to the transistor,semiconductors played a minor role in electronics research,except in detector applications for radar systems duringWorld War II. The discovery of the transistor spurredscientists and engineers to enter the exciting world ofmicroscopic behavior. As the news of the discovery spreadto the electronics industry, designers and their managers,particularly those in the military and in the emergingcomputer industry, wanted more information on how thesedevices worked. The military had a continual need to reducepower consumption and size in its weapon systems and the

46

From Gennanium to Silicon

incipient computer industry needed an small economicalhigh-speed switch to implement digital designs.

Interest in semiconductors mounted during the early1950s when Shockley developed in 1949 the theory of p-njunctions in semiconductors and the p-n junctiontransistor.25 The junction transistor was more stable thanthe point-contact transistor; structurally, the latter was morefragile. The structure of the junction transistor was lesscomplicated and better understood than that of the pointcontact transistor and by 1956 the junction transistor hadpassed the point-contact transistor as the most commonlyused transistor type in the Bell System.26

Shockley, Morgan Sparks and Gordon K. Tealexperimentally confirmed Shockley's theory in 1950 whenthey produced a number of germanium junction transistorsfor circuit analysis.27 Shockley's junction transistor went onto become the single most important solid-state devicestructure in the 1950s. The three scientists formed thejunction during the process of growing single crystals ofgermanium. In a crystal pulling furnace, a small singlecrystal of n-type (excess of electrons) germaniumis loweredinto already molten polycrystalline germanium. The seedbarely enters the top of the melted germanium whosetemperature has been raised to a value slightly above themelting point of germanium. When the n-type crystal hasgrown to the appropriate length to form the collector regionof the transistor, a pellet ofp-type (excess of holes) materialis introduced into the melt. This quickly converts thegrowing single crystal to a p-type in order to form the baseregion. Then, a second pellet of n-type material is added toform the emitter region. The result is an n-p-n structurewhose p-type base region is about 0.5 millimeters wide.

47

Seidenberg

The single crystal is then cut into small bars so that each baris an n-p-n structure (see Appendix 2b and 2c).28

Most of the transistor research performed in the1950s focused on the junction transistor. A crucialdevelopment was General Electric Corporation's successwith the alloy junction process, an inexpensive technique formanufacturing junction transistors and diodes by fusingunder elevated temperatures a pellet containing p-typeimpurities to n-type germanium. When the crystal cools, ap-type region forms in the region under the pellet, resultingin a p-n junction.29 The alloy junction process was aninexpensive and simple method for producing germaniumtransistors and diodes and was used by most high-volumesemiconductor manufacturers in the 1950s.

Shockley's theory led to a better understanding ofwhat occurs at the interfaces of metals and semiconductors.It demonstrated that p-n junctions can replace the wirecontacts of a point-contact transistor to obtain transistoraction. In addition, Shockley's work on the p-n junctionaffected the manufacturing process for the point-contacttransistor. D.K. Wilson, a BTL development engineer,observed that a pulse or series of pulses in the forwarddirection improved dramatically the electrical properties of n

type germanium point-:contact transistors and diodes. Itappeared that these electrical pulses heated the immediatearea under the metal contact to form a p-type region resultingin a p-n junction.3o

The advances made in understanding surfacecharacteristics generated renewed interest in the body of thesemiconductor material. Scientists had long understood thatuncontrolled impurities in materials tended to mask thelatter's true properties; revealing the intrinsic nature of a

48


material depended upon removing its impurities. Thepurification of germanium and silicon was based upon anancient method of purifying a crystalline solid throughrepeated fractional crystallization until 1951, when a BTLmetallurgist discovered a new approach. Using a zonemelting technique, W.G. Pfann was able to purifygermanium to a level unattainable in silicon or any othersemiconductor material,31 but zone-melting was not aseffective in purifying silicon because it melted at a highertemperature than germanium. In addition, the containersused in the purification process held impurities which easilycontaminated silicon, a more chemically active material thangermanium.

Besides purity, it was found that germanium andsilicon offered better semiconductor action when theirnormal polycrystalline structures were grown into a singlecrystal. Impurities tend to diffuse more rapidly along theboundaries that separate crystals of different orientation in apolycrystalline aggregate than through the individual crystalsthemselves. When the transistor was invented, twolaboratory methods for growing single crystals wereavailable. With the introduction of high-puritypolycrystalline germanium, BTL researchers adapted thesesingle-crystal growing techniques for germanium. At thesame time that Pfann was developing the zone-meltingprocess for purification, G.K. Teal and I.B. Little, aftergrowing their first single crystal in 1948, routinely grewsingle-crystal germanium.32 The feat of these two BTLsolid-state chemists demonstrated the importance of singlecrystal germanium in the manufacture of reliable transistorsin high volume. A high-purity single crystal has an intrinsicelectrical conductivity, affected only by variations in

49

Seidenberg

temperature. This electrical conductivity can be manipulatedby the controlled introduction of impurities. The impurities,rather than temperature, then control the resistivity of thematerial until a certain temperature level is reached, at whichpoint, the temperature becomes the predominant factor.Operation of the semiconductor beyond thi,s temperaturelevel results in device failure.

One would expect that by 1954, the transistor, whichhad been in production since 1951, would lead in the salesof semiconductor devices. The junction transistor wasproving itself as the solid-state amplifier which BTL hadsought as a replacement for the less reliable, powerconsuming vacuum tube. Additionally, the point-contacttransistor, the structure in which the first transistor wasdeveloped, was being used in many non-amplifierapplications such as oscillators, where the amplitude of thesignal was considerably more than the generated electricalnoise. Why, then, did the unit sales of diodes and rectifiersconsistently lead transistors annually during the 1950s? In1955, the unit sales of germanium and silicon diodes andrectifiers were 23.5 million versus 3.6 million fortransistors; in 1959 the figures were 120.1 million and 82.3million respectively.33

The reason is that engineers began to understandmore readily the properties of germanium. Manufacturingengineers raised yields resulting in an increase inperformance and reliability and a decrease in cost. Twoterminal devices (diodes and rectifiers) proved easier tomanufacture reliably for high-volume applications than threeterminal transistors, in spite of the fact that three-terminaldevices effectively isolated inputs and outputs. The resultwas that military and commercial computer manufacturers-

50


he latter being the largest user-preferred germanium diodesfor switching applications. The germanium diode led factoryunit sales of semiconductors in the 1950s. In 1954, the yearthat Texas Instruments announced the first transistorfabricated in silicon, 1.3 million germanium transistors wereproduced and factory sales of germanium diodes andrectifiers totaled 16.5 million units. 34 Although theinvention of the silicon transistor in 1954 offered circuitdesigners an opportunity to use silicon, the technology ofgermanium devices, particularly diodes, was so wellestablished that silicon could not match the reliability andyields of the germanium devices in large-scale production.Silicon was chosen mainly by the military establishmentwhen a rugged application required high-temperatureoperation (1250 C). The normal operating temperature ofgermanium devices did not exceed 700 C.

Computer designers preferred the germaniumjunction diode as well for their logic circuits because of itshigh-speed switching capability. Diodes designed forswitching allow practically no current to pass until athreshold voltage is reached. At that point, a large current isallowed to pass, energizing the circuit. Diode resistancedrops sharply and becomes relatively independent of thevoltage across the device. The diode functions as anelectronic switch, faster and longer lasting than anymechanical or electromechanical switch. Germaniumswitching diodes also exhibited a lower threshold voltagethan silicon devices when turned on, a useful property forlogic designers trying to minimize voltage drops.

The circuit designer in the 1950s could select fromseveral types of logic circuits. In systems where a largenumber of circuits were employed, diode logic was chosen

51

Seidenberg

because the germanium diode was relatively inexpensive,reliable and small. It was a fast electronic switch andoperated at low pqwer levels. For amplification, a functionthat diodes could not perform, engineers employed smallnumbers of transistors or vacuum tubes.

It was not uncommon to find large numbers ofgermanium diodes in a single computer system. A typicalsystem might contain close to 400,000 componentsincluding not only diodes and transistors, but alsocapacitors, resistors and coils. Germanium diodesdominated the system, comprising about 40% of thecomponent population, outnumbering even resistors, themost commonly used component in other electronicequipment. 35 The transistors were used as invertingamplifiers to compensate for circuit losses. Engineerspreferred silicon diodes where pure rectification was thegoal, typically in power supplies where higher voltages andtemperatures dominated. Circuit designers selectedtransistor logic primarily for control applications wherespeed was not essential. Resistor-transistor logic (RTL),although relatively slow, is easier to design because itcontains only one active device. A typical RTL circuitcontained three resistors which performed the logicoperation, and a transistor which provided an output voltagelevel opposite to the input level. On the other hand, diodelogic could contain up to five diodes. The number oftransistors used in control systems in no way matched thehuge amounts of germanium diodes employed in dataprocessing systems.

By 1959 germanium dominated electronic circuitsand the worldwide electronics industry was consumingmore germanium diodes than any other type semiconductor.

52


Design engineers selected the germanium diode for mostswitching circuits in military and commercial computersystems where high-temperature operation was notabsolutely essential. The annual factory sales of germaniumdiodes and rectifiers in 1959 had reached over 100 millionunits compared to 4.8 million silicon transistors. In 1959the average selling price of a silicon transistor was $14.53,compared to $1.96 for a germanium transistor. The averageprice for a silicon two-terminal device, diode and rectifier,was about four times that of a germanium unit, $1.89 versus$0.48.36 Why, then, did logic designers abruptly terminatenew designs in germanium in 1960? The answer lies at thesurface of the transistor.

By the mid-1930s, most semiconductor researchershad accepted the premise that rectification was not a propertyof the body of the semiconductor material, but a surfaceeffect. When it became clear that rectification occurred at thesurface, researchers urgently sought a scientificexplanation. 37 One began to emerge in the wake of theintense efforts in applied science that were occasioned byWorld War II.

Because germanium and silicon were elementalmaterials and relatively simple to understand in terms ofquantum mechanics, most of the semiconductor researchwas concentrated on these two materials. After the war,researchers at BTL preferred to work with germaniumbecause it proved to be a more tractable material than silicon.Germanium could be manipulated at lower temperatures andwas not easily contaminated. However, as device structuresdiminished in size, it became apparent to the research staff atBTL that silicon was the material of the future. 38 Itsintrinsically higher energy gap meant that silicon devices

53

Seidenberg

could operate at a significantly higher ambient temperaturethan could germanium. At a given temperature, the rate ofthermally generated charge carriers are lower in silicon thanin germanium because the former requires more energy tomove an electron from the valence band to the conductionband, resulting in lower leakage currents. In addition,silicon has a higher thermal conductivity, allowing heat toflow away from the junction at a much faster rate in siliconthan in germanium.

Thus, by the early 1950s, Jack A. Morton, whoheaded up transistor development at BTL, and who workedvery closely with the BTL research staff and those who ranthe transistor manufacturing operations at Western Electric,began shifting development activities to silicon. However, itwas not easy to transfer scientific advances in silicon made atBTL to the production lines at Allentown, PA. Siliconproved to be more chemically active and more easilycontaminated than germanium, causing low yields in eachstep of the manufacturing process. Additionally, the highmelting point of silicon, higher than that of germanium,created difficulties in controlling the process. When WilliamPietenpol, a Morton successor, left BTL in 1958, theAllentown facility was not producing any silicon transistors.Even though silicon was being used for rectifiers anddiodes, the only transistors manufactured at Allentown weremade of germanium.39 AT&T exercised a tight control oversemiconductor manufacturing operations which precludedany precipitous move into silicon transistors unless themanufacturing management had a firm control of theprocess.

During the same period, the rapid,progress made insemiconductor research at BTL prompted the elevation of

54


this activity to a departmental footing equal with physics,chemistry and metallurgy. However, because of thecontinuing advances made in all areas, these departmentallines were not neatly recognized in practice. There was aconsiderable blurring of departmental lines as researchersexchanged information both formally aQd informally on aconstant basis; Morgan Sparks remembers that friendshipand similarity of research interests tended to overridedepartmental formalities.4O

The progress made in understanding bodyphenomena in transistors and the availability of high-purity,single-crystal material contributed to an overall effort toexplain surface effects. In 1954, C.A. Lee of BTLdeveloped a process for diffusing the impurities of oneconductivity type into the surface layers of a semiconductorbody of opposite conductivity (i.e., n-type impurities into ap-type body, or p-type impurities into an n-type body) atelevated temperatures under very controlled conditions. Theimmediate advantage of this process was to increase greatlythe frequency range of the transistor.41 It was easier toaccomplish this diffusion in germanium than in silicontransistors; silicon was not available in the purity that couldbe had with germanium and the diffusion process in siliconrequired higher temperatures. Although BTL scientists in1955 did demonstrate diffusion in silicon, the yield of thesedevices was poor and their reliability was questionable.42

In the same year, L. Derick and C.l. Frosch devisedan oxide-masking technique for controlling against undesiredimpurities during the diffusion process by thermally forminga silicon dioxide mask on a wafer of silicon. This wasaccomplished by "heating the wafer to between 11000 C and14000 C in an oxidizing atmosphere to form a surface oxide

ss

Seidenberg

film ... between 1500 and 3500 Angstroms thick."43 Themask is removed from selected portions of the wafer and thedesired conductivity-type impurities are diffused into thesilicon body, where the oxide film had been removed. Theoxide film is then re-grown in water vapor and the vapor ofan impurity opposite in conductivity to the first impurity.The second impurity can be diffused through the silicondioxide layer to form a diffused transistor junction. Forinstance, into a p-type material, phosphorus, an n-typeimpurity, is diffused through the windows in the oxidemask. The silicon dioxide film is reformed and boron, a ptype impurity, is introduced to form a p -n -p diffusedjunction transistor. Diffusion, a controlled high-temperatureprocess at the atomic level, permitted a more preciseconcentration of desired impurities.

This technique of using an oxide mask to shut outundesired impurities was recognized at BTL as an importantstep in transistor manufacture.44 However, it did not.directly address the problem of surface effects and itremained for another BTL research team to demonstratesuccessfully the stabilization of a silicon surface. In 1957M.M. Atalla and his group at BTL stabilized a silicon surfaceby thermally growing on it a silicon dioxide film. They hadbeen systematically investigating the silicon/silicon dioxideinterface and determined that slow states at the siliconsurface were eliminated by the oxide film. The groupcautioned that "the oxidation process not only provides astable surface but a completely new kind of surface with itsown characteristic properties."45 Atalla and his group hadused silicon dioxide, chemically identical to quartz, topassivate the silicon surface. A very thin film of silicondioxide does grow naturally on a clean ,surface of silicon,protectin~ the surface from corrosion much like that of a

56


naturally occurring copper oxide on copper but the naturalsilicon dioxide film is too thin to give chemical protection orto act as an insulator. However, by thermally growing athicker film, Atalla and his group not only protected thesurface from external contamination, but also effectivelyreduced the density of surface states at the silicon/silicondioxide interface by several orders of magnitude.

Meanwhile, no progress was being made on thesurface stabilization of germanium and BTL had shifted itsemphasis to silicon by the mid-1950s. None of the membersof the former BTL staff interviewed for this chapterremembered any specific work being performed to form anoxide of germanium. Outside of BTL, the major germaniumproducers were not committed to use money or personnel forserious applied research on surface effects in germanium.Nick Calandrello, a development engineer with CleviteTransistor Products, a major germanium devicemanufacturer in the 1950s, recalls no program to grow oxidefilms on germanium. John Royan, a research engineer atTransitron, a leading germanium diode producer, workedunsuccessfully in the late 1950s to produce a stablegermanium oxide. In 1960, he was hired by NationalTransistor Corporation, a newly formed firm that hadconcerns, later found to be baseless, that Transitron wasachieving some success with germanium oxide films.46

James Battey, a physicist with Sylvania Electric, a leadingtube manufacturer making germanium devices, recalls noprogram for the oxide passivation of germanium.47

Germanium device manufacturers were aware of the benefitsof thermally grown oxide films but were preparing to meetthe threat by manufacturing silicon devices because theywere unable to develop a stable oxide film for germanium.

In 1959, Jean Hoerni of Fairchild SemiconductorCorporation implemented in production the oxide-

57

Seidenberg

passivation techniques of Atalla's group. An n-p-n planarpassivated silicon transistor was fabricated using keyprocessing techniques first discovered at BTL. The oxidemask for diffusion was based upon the work of Derick andFrosch, although the data on mask thickness, diffusion timesand temperature were derived from the first modeling of theoxide mask by C.-T. Sah and his colleagues at Fairchild.48

Hoerni's device left "permanently in place the coatingcovering all parts of the p-n junction that extend to thesemiconductor surface. "49 Electrical contacts wereevaporated onto openings incorporated into the mask designfor the purpose of attaching leads to the inside of thetransistor package. What Hoemi had accomplished was touse BTL's process advances to produce a silicon transistorthat lent itself to batch processing, and whose yield, cost,and reliability potential seemed superior to the contemporarymethods of germanium transistor manufacture. The key tothis device was the oxide-passivation technique developedby Atalla and his group. Sah, a former employee ofShockley Semiconductor Laboratories, whose majorcontribution lay in oxide masking, agrees that the successfulattempt by the BTL group to stabilize the surface of siliconwas the most significant technological advance in thedevelopment of the silicon planar passivated transistor.50

All of the key processes-impurity diffusion, oxidemasking, and surface passivation-were developed at BTL.Even the planar structure was not a new concept, althoughthis was the first time it had been used successfully intransistor production. The planar diode, a junction devicewhose electrodes lie in the same plane, already had beendescribed by a BTL researcher in 1954.51 The planartransistor is a flat-surfaced structure in which all electrically

58


active regions of the device, emitter, base and collector,terminate in the same plane. This structural planarity wasadvantageous for oxide masking and oxide passivation(Appendix 2d). Hoerni's feat was to put all these elementstogether to produce a transistor whose merits were eminentlysaleable to logic designers looking to incorporate switchingand amplification functions into one device.

Fairchild Semiconductor proved to be the rightmessenger for the new product. The company was foundedin 1957 by eight engineers from Shockley SemiconductorLaboratories in Palo Alto, California. Shockley had leftBTL in 1954 to capitalize on his achievements. He returnedto his hometown to form a semiconductor research companyin 1956 where he found little difficulty in attracting youngand talented engineers. However, these engineers soonbecame unhappy with Shockley's managerial style. Onequirk which particularly gained prominence was Shockley'spenchant for publishing the salaries of his employees on thecompany bulletin board. They decided to leave the companyen masse and form their own company

The group obtaining financial backing in 1957 fromFairchild Camera and Instrument Corporation. Logicdesigners were anxious to sample the planar-passivatedtransistor. Fairchild Semiconductor was able to providedevices which passed the customers' qualifying tests. Bythe end of 1960, practically all computer logic designers hadswitched from germanium diodes to silicon transistors anddiodes. The development of the planar process provided thesemiconductor industry with the capability of massproducing reliable miniaturized high-performance silicondevices whose switching speeds, rectification efficiencies,breakdown voltages, and power-dissipation ratings weresuperior to those of germanium.

59

Seidenberg

The shift from germanium to silicon signaled adramatic reduction in the price of silicon transistors. In 1960the average selling price of a germanium transistor was$1.70, compared to $11.27 for a silicon transistor. Fiveyears later, the average unit price of a germanium transistorhad dropped to $0.50, but the silicon transistor haddecreased more precipitately to $0.76. During the sameperiod, the unit sales of germanium transistors had risenfrom 119 million to 334 million. In contrast, the unit salesof silicon transistors during the same time had leaped from 9million to 273 million.52

The silicon transistor was one of the few innovativedevices not introduced by BTL during the 1950s. However,as Pietenpol has pointed out, the policy of BTL to disclosepublicly its progress in transistor research and technologyenabled companies such as Texas Instruments, Inc.,inventor of the silicon transistor, and Fairchild to developsemiconductor expertise. In the early 1950s, BTL heldseveral disclosure symposia, granting patent licenses undernon-exclusive and cross-licensing conditions at veryreasonable costs to all comers. Many small semiconductordevice companies started manufacturing transistors withoutobtaining patent licenses, but BTL generally ignored thesetransgressions. William Pietenpol attributes this open policyto Jack Morton. Sometime after the public announcement ofthe transistor in 1948, Mervin Kelly instructed Morton tofind an answer to "what should be done with the inventionof the transistor."53 When Kelly returned, after a month inEurope, Morton advised him that the best policy would be todisseminate information about the transistor in as broad amanner as practicable, both nationally and internationally.Holding that BTL's primary focus should remain on

60


telephone technology, Morton recommended that the bestway to advance BTL's interests would be to publish widelyand cross-license all interested parties, includingcompetitors. In this way, BTL could be certain that alladvances in transistor technology made by outsiders wouldflow rapidly into the Bell System. Of course, there are otherreasons which could be attributed to this policy of opendisclosure. Many of the papers published in the BellSystem Technical Journal indicate support by the DefenseDepartment under specific contracts. The DefenseDepartment encouraged dissemination in order to rapidlyestablish an industrial base for the transistor. AT&T alsowas concerned with anti-monopoly accusations. However,Pietenpol, who later became the director of development forsolid-state devices at BTL, believes that it was Morton'svision of how AT&T could best benefit from the inventionof the transistor that influenced top management to set thepolicy for disclosure.

Although the full disclosure policy of BTL helpedshape the rapid growth of the semiconductor industry in the1950s, five years elapsed after the development of the siliconpassivated planar transistor before silicon surpassedgermanium as the dominant starting semiconductor materialin production. The reason lies primarily in the dynamics ofthe business practices of the semiconductor industry'slargest customer, the computer industry. Between 1957 and1959, computer manufacturers expanded the production ofthe first generation of electronic computers. In order torecover their costs and accrue a profit sufficient to design thenext generation of computers, many managers in thecomputer industry estimated that five years of systemproduction were required in order to obtain an acceptable

61

Seidenberg

return on investment. Kenneth Flamm has suggested thattwo or three years were needed for a return on research-anddevelopment investment plus another two or three years "forsignificant technological rents."54 This analysis is supportedby actual sales of germanium diodes between 1959 and1965. In 1959, unit sales of germanium two-terminaldevices were 66.5 million, while in 1965 they had increasedsix-fold to 383.1 million; most of these devices weregermanium gold-bonded switching diodes.55

Not only did computer manufacturers need time torecoup their investment, but silicon device vendors needed atleast two of these years to negotiate specifications, estimatecosts and pricing, submit lots for qualification testing, andrenegotiate specifications and prices. Submission formilitary qualification generally took more than a year longer.During this time production sites were selected, facilitiesbuilt, and factories manned. Manufacturers of germaniumlogic devices encouraged the delay in switching over tosilicon by dropping prices as cost reductions were realizedfrom their high-volume manufacturing experience. Thesesuppliers continued to increase yields and refinemanufacturing processes to a level where the average sellingprice of a germanium logic diode fell below ten cents, aboutone-quarter to one half-lower than that of a silicon diode,and at least five time less expensive than a silicon switchingtransistor.56

Some questions still remain: Was it inevitable thatsilicon would replace germanium? Was silicon an inherentlysuperior material? The history of the virtual replacement ofgermanium by silicon underlines the significance of bothmaterial properties and technical processes in the newspecialization of solid-state electronics. The research and

62


development efforts of professionals who select their fieldsof investigation based upon technical interests played a largerole in overturning germanium in favor of silicon, inparticular the BTL researchers who chose to study thesurface effects of silicon rather than germanium. Efforts byother corporate research and development groups topassivate the surface of germanium appear not to have beensustained by the same quality and quantity of scientistsworking on silicon at BTL. No one to date has grown astable oxide passivation layer on germanium, in spite of thefact that in the 1950s many companies were totallydependent upon germanium devices for their sales andprofits. It was not inevitable that silicon would replacegermanium. Silicon was developed by BTL researchersbecause they perceived that its material properties were moreuseful for telecommunications than those of germanium.

Morgan Sparks recalls that the reason silicon waschosen for investigation over germanium was because theforbidden energy gap of germanium is smaller than that ofsilicon. The smaller germanium gap between its conductionband minima and valence-band maxima meant that theleakage currents of germanium would be higher than thoseof silicon, permitting the latter to operate safely atsignificantly higher temperatures than the former.

Why did the BTL scientists not select a parameter ofgermanium which was inherently more useful to that ofsilicon as guidepost for research? Electron and holemobilities in germanium are considerably higher than insilicon, making the former a much faster switching material,but none of the BTL researchers interviewed mentionedswitching speed as a factor in favor of continuing researchon germanium; most mentioned the superior operating

63

Seidenberg

temperature of silicon, the implication being that withinreasonable application limits, reliability was preferable toperformance. Another issue, unexplored here, is thatswitching speed is presumably a parameter more useful incomputer logic than in telephone amplification.

Eventually, the inherent performance advantages ofgermanium in switching circuits were overcome by theintensive investigations of silicon by BTL and the rest of theindustry. Silicon had surpassed germanium in switchingspeeds by the beginning of the 1960s. Developers shortenedtransit times by reducing the feature dimensions of thetransistor. Germanium, as its dimensions were reduced,became more difficult to handle because the material had apoorer thermal conductivity than silicon; heat flowed awayfrom the germanium p-n junction at one third the rate ofsilicon.

Silicon replaced germanium through the combinedeffect of a variety of researchers, properties, and processesall existing in an environment that was dominated by BTL.What happened there supports the belief that "science,although part of a larger social and economic context, isultimately driven by individual desire and deterrnination."57The scientists and engineers at BTL wanted to participate onthe leading edge of the new science of the solid state, andquantum mechanics led scientists to investigate themicroscopic and unseen, where the behavior of the electronin solids was a mystery to be unraveled and anyone whosucceeded in doing so could gain professional recognition.Furthermore, researchers and engineers themselves chosethe materials to be investigated. Their decisions were basedupon the perception that the properties of elementalsemiconductors would be easier to understand and

64


manipulate than the pre-war compound materials. Becausethe research and development of semiconductors is aninterdisciplinary endeavor, progress was neededsimultaneously in several areas.

This chapter has touched on other elements of theBTL environment. The overarching objective at BTL was toreplace a mechanical telephone system with an electronicone. At a time when science and technology seemed to bethe answer for how humans could improve their lot in theworld, BTL started its semiconductor research activities longbefore its competitors and was able to gather a group ofscientists wh.o could attract the most talented newcomers toits staff after World War II.

Semiconductor device development can be comparedto system development. Although often invisible to thenaked eye, the development of a device suffers from thesame ebbs and flows more apparent in large systems. In thebeginning, the scientists at BTL found germanium easier tohandle than silicon. They could purify germanium moreeasily. The higher processing temperatures required forsilicon because of its inherent properties created complexitiesin handling the material and measuring results. Earlytransistor structures were large enough to neutralize thelimited heat-handling capabilities of germanium. Asprocessing and measuring techniques became moresophisticated, smaller transistor structures could beinvestigated. Miniaturization to the scientist meantconservation of energy. It coincided with the performanceand economic goals of both the military and industry. AsBTL scientists and engineers learned to process smallerstructures in volume, the properties of silicon contributing tohigh-density, high-temperature operation pre-empted the

65

Seidenberg

performance advantages of germanium. BTL ceasedmeaningful research on germanium by the mid-1950s and,by 1960, the continuing investigation of silicon had resultedin its acceptance by system designers as the dominantsemiconductor starting material. This acceptance happenedvery quickly because of the recognition by the industry thatthe oxide passivation of silicon had created a different andapparently superior semiconductor device. However, morethan five years passed before design acceptance could betranslated into the dominance of silicon in the marketplace.During these five years, BTL's research achievementsspurred other major technological innovations including themonolithic integrated circuit and the metal-oxidesemiconductor field-effect transistor, both made practicableby oxide passivation.

66


APPENDIX 1

MILESTONES IN TRANSISTOR RESEARCHAND DEVELOPMENT

(1947.1959)

INNOVATION FIRM DATE

Transistor BTL 1947Single Crystal Germanium BTL 1948Grown Junction Transistor BTL 1950Zone Refining BTL 1951Alloy Junction Transistor GE 1951Silicon Junction Transistor TI 1954Diffusion Process BTL 1954Oxide Masking BTL 1955Oxide Passivation BTL 1957Planar Process Fairchild 1959

Compiled by Author

67

Seidenberg

APPENDIX 2

EMITTERPIN

IN$ULATINGPLUG ......

CONTACTWIRES ' ... ....

COLL.ECTORPIN

.sEeo__'_-"'i' 7 ENe

la. Point Contact Transistor(Shive: 177]

1b. n-p-n Grown Junction Crysta1[Shi ve: 188]

p collector

basecontact

si 1icon --,*---:"'-l"I~~~~+--fIIl--\dioxide

emitterjunction

collectorjunction

EMITTER COl.LECTORJUNCTI~ 8~SE J~NCTION

1c. n-p-n Junction Transistor

[Shive: 188]

1d. Planar Transistor 1960

(Diagram by Author)

68

2

3

4

5

6

7


S.M. Sze, Semiconductor Devices: PioneeringPapers (Teaneck, NJ: World Scientific Publishing Co.,1991).

T.J. Misa, "Military Needs, Commercial Realities, andthe Development of the Transistor, 1948-1958," inMilitary Enterprise and Technological Change, ed.M.R. Smith (Cambridge, MA: The MIT Press, 1987),284.

W. Skinner and D.C. Rogers, Manufacturing Policyin the Electronic Industry: A Casebook ofMajorProduct Problems, 3rd ed. (Homewood, IL: RichardD. Irwin, Inc., 1968), tables, 76-79. These tables wereoriginally produced by the Electronic IndustriesAssociation (EIA).

E. Braun and S. Macdonald, Revolution in Miniature(New York: Cambridge University Press, 1978), 7374.

P.R. Morris, A History of the World SemiconductorIndustry (London: Peter Peregrinus, Ltd., 1990), 68.

J.E. Tilton, International Diffusion of Technology:The Case ofSemiconductors (Washington, DC: TheBrookings Institution, 1971), 65-68.

S. Millman, ed., A History of Engineering andScience in the Bell System: Physical Sciences, 19251980 (AT&T Bell Telephone Laboratories, 1983); andF.M. Smits, A History of Engineering and Science inthe Bell System: Electronics Technology, 1925-1975(AT&T Bell Laboratories, 1985). See also A. Many,

69

Seidenberg

Y. Goldstein, and N.B. Grover, SemiconductorSurfaces (New York: North-Holland PublishingCompany, 1965). This is an excellent technical workon semiconductors surfaces that provides an organizedview of the rapidly expanding field of surface researchin semiconductors and offers useful references andnotes for the historian.

8

9

Misa, 253-287.

L~H. Hoddeson, "The Roots of Solid-State Research atBell Labs," Physics Today (March 1977): 27..

10 W. Shockley, Bell Laboratories Record 50 (December1972): 340.

11 J.C. Williams, "Frederick E. Terman and the Rise ofSilicon Valley," in Technology in America, ed. C.W.Purcell, Jr. (Cambridge, MA: The MIT Press, 1990),284.

12 C. Weiner, "How the Transistor Emerged," IEEESpectrum 10, no. 1 (January 1973): 32.

13M. Eckert and H. Schubert, Crystals, Electrons,Transistors: From Scholar's Study to IndustrialResearch (New York: American Institute of Physics,1986), 117-118.

14 George F. Dacey, interview by Philip Seidenberg, 9August 1994.

15 Morgan Sparks, interview by Philip Seidenberg, 10August 1994.

16 C.G.B. Garrett, interview by Philip Seidenberg, 8August 1994.

70


17 Skinner and Rogers, 76-79.

18 G.L. Pearson and W.H. Brattain, "History ofSemiconductor Research," Proceedings of theInstitute of Radio Engineers 43 (December 1955):1797.

19 Shockley, 340-341.

20 J. Bardeen, Bell Laboratories Record 50 (December1972): 336-337.

21 Bardeen, 337.

22 W. Shockley, "On the Surface States Associated with aPeriodic Potential," Physical Review 56 (August 15,1939): 317-323.

23 J. Bardeen, "Surface States and Rectification at a Metalto Semiconductor Contact," Physical Review 71 (May15, 1947): 717.

24 J. Bardeen and W.H. Brattain, "The Transistor, aSemiconductor Triode," Physical Review 74 (July 15,1948): 230-231.

25 W. Shockley, "The Theory of p-n JunctionTransistors," The Bell System Technical Journal 28(July 1949): 435-489.

26 N.J. Herbert, "Point Contact Transistor Development,"Bell Laboratories Record 34 (February 1956): 46-50.

27 W. Shockley, M. Sparks, and G.K. Teal, "p-nJunction Transistors," Physical Review 83 (July 1,1951): 151-162.

28 J.N. Shive, The Properties, Physics, and Design of

71

Seidenberg

Semiconductor Devices (Princeton, NJ: D. VanNostrand Company, Inc., 1959), 187-188.

29 J.E. Saby, "Fused Impurity P-N-P JunctionTransistors," Proceedings of the IRE 40 (November1952): 1358-1360.

30 D.K. Wilson, "Semiconductor Diodes," BellLaboratories Record 33 (June 1955): 227-231.

31 W.G. Pfann, "Principles of Zone Refining,"Transactions of the American Institute of Miningand Metallurgical Engineers 194 (July 1952): 747753.

32 G.K. Teal and J.B. Little, "Growth of GermaniumSingle Crystals," Physical Review 78 (June 1950):647.



35 J.E. Schwop and R.J. Sullivan, SemiconductorReliability (Elizabeth, NJ: Engineering Publishers,1961), 254-255.


37 G.L. Pearson and W.H. Brattain, "History ofSemiconductor Research," Proceedings of theInstitute of Radio Engineers 43 (December 1955):1797-1798.

38 Sparks interview.

39 William J. Pietenpol, interview by Philip Seidenberg,14 August 1994.

72


40 Sparks interview.

41 C.A. Lee, "A High-Frequency Diffused BaseGermanium Transistor," Bell System TechnicalJournal 35 (January 1956): 23-24.

42 M. Tanenbaum and D.E. Thomas, "Diffused Emitterand Base Silicon Transistors," Bell System TechnicalJournal 35 (May 1956): 1-22.

43 U.S. Patent 2,802,760, Oxidation of SemiconductiveSurfaces for Controlled Diffusion. Filed 2 December,1955 by L. Derick and C.J. Frosch.

44 K.E. Daburlos and H.J. Patterson, "Oxide Masking,"Bell Laboratories Record 38 (November 1960): 417-420.

45 M.M. Atalla, E. Tannenbaum, and E.J. Scheibner,"Stabilization of Silicon Surfaces by Thermally GrownOxide," Bell System Technical Journal 38 (May1959): 749-783.

46 John Royan, interview by Philip Seidenberg, 29 July1994.

47 James Battey, interview by Philip Seidenberg, 18August 1994.

48 C.-T. Sah, "Evolution of the MOS Transistor -- FromConception to VLSI," Proceedings of the Institute ofElectrical and Electronics Engineers 76 (October1988): 1280-1326; and C.-T. Sah, interview by PhilipSeidenberg, 8 July 1994.

49 U.S. Patent 3,025,591, Method of ManufacturingSemiconductor Devices. Filed 1 May, 1959 by J.A.

73

Seidenberg

Hoemi; and U.S. Patent 3,064,167, SemiconductorDevice. Filed 1 May 1959 by J.A. Hoemi.

50 Sah, Interview; and Sah, "Evolution of the MOSTransistor," 1290.

51 R.H. Kingston; "Switching Time in Junction Diodesand Junction Transistors," in Transistor Technology,Vol. II, ed. F.J. Biondi. (Princeton, NJ: D. VanNostrand Company, Inc., 1958), 166-179.


53 Pietenpol interview.

54 K. Flamm, Targeting the Computer: GovernmentSupport and International Competition (Washington,DC: The Brookings Institution, 1987),233-234.



57 J. Kinoshita and J. Mervis, "Science in Japan: CreatingNew Structures," Science 266 (18 November 1994):1169.

74

From Germanium to Silicon Chapter 2 A History …_1997.pdfChapter2 From Germanium to Silicon A...

Documents

Transcript of From Germanium to Silicon Chapter 2 A History …_1997.pdfChapter2 From Germanium to Silicon A...