US2817608A - Melt-quench method of making transistor devices - Google Patents

Description

Dec, 24, 1957 J. l. PANKOVE MELT-QUENCH METHOD OF MAKING TRANSISTOR DEVICES Filed May 2, 1955 P N wmm xl INVENTOR. T s c: nues I. Fammi:-

United States Patent MELT-QUENCH `METHOD F MAKING TRANSISTOR DEVICES Jacques I. Panltove, Princeton, N. J., assignor to Radio Corporation of America, a corporation of Delaware Application May 2, `1955, Serial No. 505,333 7 Claims. (Cl. 14S-1.5)

This invention relates generally, to improved methods of making semiconductor devices. More particularly, it

relates to improved methods of making filamentary type transistors having a plurality of rectifying barriers which divide the filaments electrically into a plurality of separate zones.

An object of the invention is to provide improved methods for making improved semiconductor rectifying devices.

Another object is to improve rectifying barriers in semiconductor devices.

A further object is to simplify and improve the fabricai tion of semiconductor devices.

The instant invention is an improvement over the process described by W. G. Pfann in the Journal of Metals for February 1954, pages 294 et seq. Pfann describes the formation of two spaced rectifying barriers in a germanium semiconductor body doped with two different impurities capable of imparting opposite conductivity types to the semiconductor, by first melting and then slowly refreezing a portion of the body.

According to the instant invention it has been found that unexpected improvements in the electrical properties of devices produced by melting and refreezing portions of semiconductor bodies are provided by radically changing the rate of refreezing taught by Pfann. Whereas Pfann teaches melting and subsequent slow refreezing, applicant has found unexpected advantages in melting and subsequently refreezing at a rapid rate as hereinafter described.

The semiconductor bodies treated according to the nvention to form transistor devices are most conveniently filamentary in form. They initially include evenly distributed impurities of both donor and acceptor types. The impurities of the type present in the greater numbers are selected to have a smaller segregation coefficient in the semiconductive body than the impurities of the opposite type. According to the instant invention such a filament is melted from one end partially along its length and then quenched by rapidly refreezing it at a minimum initial cooling rate of at least 500 C. per second. The melting and quenching produces a pair of closely spaced rectifying barriers within the lament closely adjacent to the interface between the molten and solid portions of the filament.

The invention will be explained in greater detail in connection with the accompanying drawing of which:

Figure 1 is a schematic, cross-sectional, elevational view of a semiconductor filament being processed according to the invention; and

Figure 2 is a schematic, cross-sectional, elevational view of a semiconductor filament after treatment according to the invention to form it into a transistor; and

Figure 3 is a line graph describing the distribution of donor and acceptor impurities along the length of the filament shown in Figure 2.

Similar reference characters are applied to similar elements throughout the drawing.

The process of the invention makes use of the resegre gation effect. This effect is explained in detail in the article heretofore referred to and in an article, also by Pfann, in the Journal of Metals, p. 861, for August 1952.

The segregation effect concerns the relative solubilities of solutes in the molten and liquid phases respectively of a freezing solvent. In germanium and silicon, for eX- ample, most solutes are more soluble in the liquid phase than the solid. The quantitative expression for the effect is called the segregation coeicient. The segregation coeliicient of an impurity substance in germanium, for example, may be defined as the ratio of the concentration of the impurity substance on the solid side of the interface of a growing ingot of freezing germanium to the concentration on the liquid side of the interface. (Concentration in solid concentration in liquid.) When two different impurities are present in a freezing system' in relatively small proportions they act substantially independently of each other with respect to the segregation effect.

Referring now to Figure 1, a transistor device may be made according to a preferred embodiment of the invention utilizing a filament 2 of n-type single crystal semiconductive germanium. The filament may be, for example, about 0.10 long and about .03 in diameter. It includes both donor and acceptor impurities, the donor, or n-type impurities being present in larger quantity than the acceptor, or p-type impurities. Further, the donor impurities are selected to have a substantially smaller segregation coefficient in germanium than the acceptor impurities. The impurities may consist of dispersed atoms of respectively, gallium (an acceptor impurity having a segregation coefficient of 0.1) and antimony (a donor impurity having a segregation coefficient of .005). The gallium may be present in a quantity of about 4 101'5 atoms per cc. in the germanium and the antimony may be present in a quantity of about 5 1016 atoms per cc. Both impurities are substantially uniformly distributed throughout the filament. The antimony, therefore, is the predominant impurity and determines the conductivity type of the germanium.

The filament is held vertically in a well in a graphite or ystainless steel block 4 which has a relatively large thermal capacity and acts as a heat sink. One end of the filament is brought into contact with a flat carbon or tantalum wire 6 in a non-oxidizing atmosphere such` tric current to flow through the wire to heat it to aboutv 2200 to 2800o C. The current llow is maintained for about one second during which time heat is transferred from the wire to the germanium filament to melt approximately half the length of the filament. The supporting block 4 cools the bottom portion of the filament which is remote from the wire and prevents the wire from melt-` ing the entire filament. After about onesecond the switch is opened and the filament is allowed to cool rapidly down to room temperature. The initial rate of cooling is about l000 C. per second.

The above described process produces a lilamentary body with two closely spaced p-nrectifying junctions A and B disposed transversely across the filament as shown in Figure 2. Both the junctions are formed in relatively abrupt transition zones, or regions between the n-type and p-type portions of the body. One junction (A) is formed at the surface corresponding to the interface between the molten and solid portions of the :filament at the point of its farthest advance. This junction is similar to one of the junctions described by Pfann. A second rectifying junction (B) is formed closely adjacent to the first and substantially parallel to it.

The filament is thus divided electrically into three separate regions. The end portions of the filament are of n-type conductivity and a central, relatively narrow portion is of p-type conductivity.

Pl`he filament may then be etched according to any of several known techniques to delineate the central, ptype conductivity region upon its surface. Electrical leads 2t), 22 and 24. may then be attached to the respective regions of the device and the device may be conventionally mounted and potted.

The central p-type conductivity region becomes the base region whenthe device is utilized as an n-p-n transistor. It is relatively flat and its thickness may be less than .001". This .l central region can be conveniently located at about the middle of the filament. An electrical connection forming the, transistor base connection may be made to this region` by any conventional process such as by surface alloyingran indium electrode or by the goldbonding technique utilizing goldv alloyed with4 an acceptor impurity. Suchy connections are substantially non-rectifying, orohmic with respect to the p-type conductivity region. They` form p-,n rectifying barriers, however, upon 1r-type conductivity material. Thusrif the junction between` a goldbonded lead and the filament extends beyond the edge of the p-type region and con: tacts the n-type region it does not electrically short-circuit the barrier between the regions. The only effect of such an extension is a corresponding extension, or deformation of the base region to include the entire contact area of the gold bonded lead.

The leads 24 and 2f) connected to the ntype end portions of the filament may be soldered thereto by, for example, a tin-lead` solder with which is alloyed a small percentage of a donor impurity such as antimony.

The line graph of Figure 3 illustrates the resegregation effect in the practice of the invention with respect to the embodiment heretofore described. The two lines 26 and ZSArepresent respectively, the donor, or n-type impurity concentration and the acceptor, or p-type impurity concentration at consecutive planes along the length of the lament. The origin of the graph corresponds to the end of the filament seated in the heat sink'.

During the process the impurities in the solid, unmelted portion of the filament remain substantially unchanged and in their original concentrations. Here the donor impurities predominate and the filament is of ntype conductivity. The lfilament is melted during the processnfrom its Itop ,down to the plane represented by, the pointAl inthe graph.l VJhenthe` heat supplied to they filamentris cut off 'there is initially a relatively brief time during whichthe filament cools and freezeslongitudinally at alrelatively vslovvrate such ythat the resegregation effect.

is operative. This brief time is believed to be closely related to the` time necessary to cool the molten portion of the filament to a temperature Close to its freezing point. After this brief time the remainder of the filament freezes so rapidly that the impurities present in the freezing portion are not resegregated but vare quenched 1n position.

Most of the heatdissipation during cooling is by conduction through the solid portion of the lament and into the heat sink. A consideration of the parameters involved reveals that relatively little heat is lost by radiation and by gas convection directly from the filament into the surroundingspace.` There results, therefore, a thermal gradient extendingalong the length of the filament from the heatedendtoward the heat sink. rfhis gradient induces longitudinal freezingwhich is believed to be a primary factor in causing the filament torefreeze in single crystal form` determined by the crystal lattice orientation of the non-meltediportion.

During the-time that the resegregation effect is oper ative, a portion of the filament corresponding to the increment A to B inthe line` graphie frozen. The rela-` tive concentrations of the two impurities in this region are aected by their respective segregation characteristics. Since the donor, or n-type impurity has a relatively small segregation coeiiicient, relatively small proportions of it are included in this region. rl`he acceptor, or p-type impurity, on the other hand, has a relatively large segregation coefficient and relatively large propor- Cd 0a where:

ka Y is; the segregation` coefficient. of i the acceptor impurity kd is the segregation coefficient of the donor impurity Cais the concentration of the acceptor impurity Cdis the concentration of the donor impurity For optimumrelectrical characteristics it is also desirable that ka be asmuch greater than kd as possible. This permits a maximum difference in conductivities between the end portions of the filament and the resegregated portion, which difference is conducive to desirable transistor characteristics.

By suitably selecting the impurities initially present in the filament and their respective concentrations the resistivities of the different regions after processing may be controlled within relatively Wide limits. For example, the filament may initially include a relatively low concentration of boron, a donor impurity having a large segregation coefficient, and a relatively high concentration of bismuth, an acceptor impurity having a small segregation coefficient. The relative concentrations of these two impurities may be selected to provide a relatively low resistivity in the end portions of the filament by making the bismuth concentration greatly exceed the boron concentration. Simultaneously, the resegregated, central portion of the filament is caused to have a relatively high resistivity since after resegregation the boron concen tration exceeds the bimuth concentration by only a small margin.

After the resegregated portion of the lament is frozen,V

the 'remainder'of the filament freezes so rapidlythat the resegregation effect is ineffective and the impurities are present* throughout` the remainder of the filament in substantially their original concentrations. In the end portions of the filament Vwhere the donor impurity concentration exceeds lthe acceptor impurity concentration, thc larnent exhibits n-type conductivity, and in that portion ofthe filament where the acceptor concentration exceeds the donor concentration, the filament exhibits p-type conductivity. The respective consecutive portions of the filament are electrically separatedv by p-n rectifying barriers.

One important advantage of the invention is the provision of two closely spaced p-n rectifying junctions in a :semiconductive` filament, each of which junction lies within a Vrelatively abrupt transition zone between a ptype conductivity region and an n-type conductivity region. symmetrical transistor for puse in devices such as electronic switches and reversible amplifiers.

Another advantage` is the relative ease andrapidity withv which transistor devices may be made. The entire process takes no more than a few seconds. Further, the processeof the invention is not subject to critical tolerance limitations. For example, the lengthof the filament that needbe ,melted` isnot critical since this factor affects pri- Such a structure is particularly desirable as a.

marily onlythe specific location along the length of the lament at which the two rectifying junctions are formed. The temperature of heating, also, is not critical since the melting process appears to attain thermal equilibrium. The length of the portion of the filament that is melted is determined by the amount of heat provided, while the heat sink maintains one end of the filament at a temperature well below its melting point and prevents the melting of the entire filament.

The time factor is believed to be primarily responsible for the lack of deformation of the melted portion of the filament. If the heat is applied too long while the material is melted, surface tension forces may cause the molten material to ballup at the end of the filament. In this case the cooling part of the process is less uniform and non-planar or non-uniform rectifying barriers may be formed. It is preferred, therefore, to start cooling immediately after the heated end of the filament attains the desired temperature of 1500 to 2000 C.

In making practicable devices the principal critical factor appears to be the spacing between the two junctions. The spacing is affected primarily by the rate of cooling of the device after the heating is removed. A rapid rate of cooling provides closely spaced junctions, whereas a relatively slow rate of cooling provides relatively widely spaced junctions. In either case, however, the rate of cooling is extremely rapid when judged by the rates of cooling previously employed in the manufacture of semiconductor devices and materials.

One way to control the rate of cooling is to Vary the temperature of the heat sink to control the thermal gradient along the filament. It has been found that convection and radiation heat losses are of only minor irnportance in determining the rate of cooling of the filament. The principal heat loss is by conduction through the filament to the heat sink. The heat sink may be cooled, for example, by directing a stream of cool water or liquid nitrogen through it to increase the thermal gradient along the filament, thereby to increase the cooling rate to minimize the spacing between the two junctions.

In order to ensure an abrupt transition zone at the second formed junctions, it is necessary to quench the end portion of the filament. Otherwise, the resegregation effect proceeds along the length of the filament as shown by Pfann and produces a gradual transition zone. The quenching is also highly desirable in order to provide close spacing between the two junctions without placing both of them critically close to one end of the filament.

If the filament is melted and subsequently slowly refrozen, it has been found that either the two junctions must be widely spaced or formed very close to the melted end of the filament. If the two junctions are formed by slow freezing sufficiently close together to provide desirable electrical characteristics at relatively high frequencies they must be formed so close to the end of the filament that any contact made to the end of the filament is apt to damage or to interfere with the junction adjacent thereto. According to the invention, filamentary transistors may be made having pairs of p-n rectifying junctions closely spaced andextending transversely across the filaments at any desired locations along the lengths of the filaments. l

It has previously been thought that the growth of a single crystal of semiconductive materials such as germanium and silicon depended to a large extent upon a slow rate of cooling. According to the instant invention, however, it has been unexpectedly discovered that even when quenched at the extremely rapid cooling rates heretofore described, partially melted filaments of semiconductive germanium and silicon refreeze in substantially single crystal form.

To provide uniform characteristics among successive devicesjand to provide` optimum' characteristics in semiconductor devices such as transistors generally, it has been found that each of the devices must comprise a single crystal of semiconductive material. llf any portion of a device such as the transistor illustrated in Figure 2, for example, is crystallographically inhomogeneous the electrical characteristics of the device are apt to be inferior, unpredictable and very possibly variable with time.

The practice of the invention is, of course, not limited to a specific device incorporating the specific materials heretofore described. For example, devices similar to the one shown in the drawing may be made utilizing semiconductive silicon or an alloy of silicon and germanium. Also, the particular impurities and their concentration may be varied according to the principles heretofore set forth to produce p-n-p as well as n-p-n devices.

That is, an acceptor impurity having a small segregation* coefficient may be initially included in the filament as the4 dominant impurity in conjunction with a donor impurity having a relatively large segregation coefficient. Then:

is the required relationship `between the impurities initially in the filament.

A partial list of impurities that may be used in conjunction With semiconductive germanium and silicon according to the principles of the invention is set out below. The segregation coeliicients listed are those with respect to germanium and are empirical. They are determined for a slow rate of freezing such as utilized in the freezing of the resegregated portion of a filament according to the invention. The figures given may be used as a guide in selecting significant impurities for the purpose of the invention. The respective coefficients for the listed materials in silicon are directly proportional to the coeicients in germanium.

There have thus been described improved methods of making filamentary, junction type transistors by a meltquench process in which a longitudinal portion of a filamentary crystal of a semiconductive material is melted and then rapidly refrozen to form a pair of spaced, parallel, planar p-n rectifying barriers in the crystal.

What is claimed is:

l. Method of making a semiconductor device comprising the steps of melting throughout the entire thickness a longitudinal portion of a crystalline semi-conductive filament, said filament including both acceptor and donor conductivity type-determining impurities in relative proportions such that the following relationship is satisfied:

k C. 0. t0

GII Where Cx is the atomic concentration of said impurities of one maintaining another longitudinal portion of said filament in its solid state, and subsequently quenching said melted portion by rapidly refreezng it at a minimum initial coolingr-ate of-`at least-500 C. per secondiso that solidit'ieationVA where Cxis the atomic concentration of said impurities of one type,

Cy is theatornic'concentration of said impurities o the opposite type,

kX is the segregation coefficient of said impurities of said one type in said iilament, and

ky is the segregation coeflicient of said impurities of said opposite type in said filament,

maintaining another longitudinal portion of said lament at the opposite end thereof from said one end in its solid state at a temperature below its melting point thereby to establish a thermal gradient along the length of said filament, and subsequently quenching said melted portion by rapidly extracting heat from said solid portion at an initial rate oi' at least 500 C. per second thereby to maintain said thermal gradient and to cause said melted portion to freeze longitudinally starting at the unmelted cool end and proceeding to the other melted end to form a pair olf spaced parallel rectiiying barriers in said lilament.

3. Method. of making a semiconductor device comprising the steps of melting a longitudinal portion of a single crystal semiconductive filament throughout its entire thickness by applying heat at one end of said filament, said filament including both acceptor and donor conductivity type-determining impurities in relative proportions such that the following relationship is satisfied:

where C is the atomic concentration of said impurities of one type..

Cy is atomic concentration of said impurities of the opposite type,

kx is the segregation coeflicient of said impurities of said one type in said filament, and

ky is the segregation coeflicient of said impurities of said opposite type in said filament,

maintaining another longitudinal portion of said filament at the opposite end thereof from said one end in its solid state at a temperature below its melting point thereby to establish a thermal gradient along the length of said filament, and subsequently quenching said melted portion by rapidly extracting heat from said solid portion at a rate suflicient to provide an initial cooling rate of at least 500 C. per second in the molten portion of said filament thereby to maintain said thermal gradient and to cause said melted portion to freeze longitudinally starting at therfunmelted cool end and proceeding to the other end to form a pair of spaced parallel p-n rectiiying barriers in said filament.

4. Method of making a semiconductor device comprising the steps of melting a longitudinal portion of-a single crystal semiconductive germanium filament throughoutits' entirethickness by" applyingheat 'at one end there- OLsaid'Iilament including both acceptor anddonor conlll agonistico@ s ductivity` f type-determining impurities in' relative proportions suchthat the following relationship is satisiied where Ca is the atomic concentration of said acceptor impurities,

Cd is the atomic concentration of said` donor impurities,

le, is the segregation coefficient ofsaid acceptor impurities in germanium, and

kd is the segregation coefficient of said donor impurities in germanium,

maintaining another longitudinal portion of said filament at the opposite end of said filament from said one end` in its solid `state at a temperature below its melting point, and subsequently quenching said melted portion by cooling it at a minimum initial cooling rate of 500 C. per second so that solidiication starts at the unmelted cool end andproceeds to the other end, thereby to form a pair of spaced parallel p-n rectifying barriers in said filament.

5. Method of makinga semiconductor device comprising the steps of melting a longitudinal portions of a single crystal semiconduotive. germanium filament throughout its entire thickness by applying heat at one end thereof, said filament including both acceptor and donor conductivity type-determining impurities in relative proportions such that the following relationship is satisfied:

where maintaining another longitudinal portion of said iilament at the Opposite end of said filament from said one end in its solid state at a temperature below its melting point, and subsequently quenching said melted portion by cooling it at a minimum initial cooling rate of 500 C. per second so that solidiiication starts at the unmelted cool endrandproceedsto theother end, thereby to form a pair of spaced parallel p-n rectifying barriers in said lilament.

6. Method of making a semiconductor device comprising the steps ofmelting a longitudinal portion of a single crystal semiconductive germanium lilament throughout its entire thickness by applying heat at one end of said lilanient in a non-oxidizing atmosphere, said filament including dispersed atoms of gallium and antimony, the antiinony atoms being present in a greater number than the gallium atoms, the number ot` gallium atoms being greater than 1/0 the number of antimony atoms, maintaining another longitudinal portion of said filament at the opposite end thereof from said one end in its solid state at a temperature below its melting point, and subsequently quenching said melted portion at a minimum initial cooling rate of 500 C. per second so that solidification starts at the unnielted cool end and proceeds to the other end, thereby to form a pair of spacci parallel p-n rcctiying barriers in 'said lament.

7. Method of makinga semiconductor device comprising the steps of melting a longitudinal portion of a single crystal semiconductive germanium filament throughout its entire thickness by applying heat at one end of said lament in a non-oxidizing atmosphere, saidlanient including, dispersed atoms of gallium andantimony, the antimony atoms being present in a greater number than the gallium atoms, the number of gallium atoms being greater than the number of antimony atoms, maintaining another longitudinal portion of said filament at the opposite end of said filament from said one end in its solid state at a temperature below its melting point thereby to establish a thermal gradient along the length of said filament, and subsequently quenching said melted portion at a minimum initial cooling rate of 500 C. per second by extracting heat from said solid portion thereby to maintain a thermal gradient along the length of said filament to cause said melted portion to freeze longitudinally so that solidica- 10 tion starts at the Iunmelted cool end and proceeds to the other end in single crystal form as determined by the crystal lattice of said solid portion, and to form a pair of spaced parallel pn rectifying barriers in said filament.

References Cited in the file of this patent UNITED STATES PATENTS 2,739,088 Pfann Mar. 20, 1956 FOREIGN PATENTS 1,065,523 France Ian. 13, 1954

Claims (1)

1. METHOD OF MAKING A SEMICONDUCTOR DEVICE COMPRISING THE STEPS OF MELTING THROUGHOUT THE ENTIRE THICKNESS A LONGITUDINAL PORTION OF A CRYSTALLINE SEMI-CONDUCTIVE FILAMENT, SAID FILAMENT INCLUDING BOTH ACCEPTOR AND DONOR CONDUCTIVITY TYPE-DETERMINING IMPURITIES IN RELATIVE PROPORTIONS SUCH THAT THE FOLLOWING RELATIONSHIP IS SATISFIED:

Images