US2824030A - Method of preparing semiconductive materials - Google Patents

Description

' Fgb. 18, 1958 J. w. RUTTER El'AL 0 METHOD OF PREPARING SEMICONDUCTIVE MATERIALS Filed July 21 1955 Z Sheets-Sheet 2 CofSb) POINT AT WHICH SOL lD/F/CA T/ON BEG/NS AFTER MIXING CONCENTRA T/ON Inventor JOHN WILL RU R WILL/AM ART T/LLER By: M w w Att'ys United States Patent F METHOD OF PREPARING SEMICONDUCTIVE MATERIALS John W. Rutter and William A. Tiller, Toronto, Ontario,

Canada, assignors to Canadian Patents and Development Limited, Ottawa, Ontario, Canada, a Canadian company Application July 21, 1955, Serial No. 523,597

5 Claims. (Cl. 1481.5)

This application is a continuation-in-part of our copending application Serial Number 440,462, filed June 30, 1954, now abandoned.

This invention relates to the preparation of semiconductive material by means of the controlled re-distribution of solute impurities in the solidification of suitable metals. In its more specific aspects, the invention is concerned with the preparations of germanium and like crystals from which transistors and p-n junction rectifiers inay be made.

'It is well known that the solid which is frozen from an impure liquid does not have the same composition as the liquid from which it solidifies: for instance, if a single-phase alloy is solidified under non-equilibrium conditions a redistribution of the impurity occurs and the resulting solid is not of uniform composition. In certain systems of this kind, the solid formed contains less impurity than the liquid from which it forms, which means that a certain proportion of the impurity in the liquid part of the melt is rejected on solidification of part of the liquid into the remaining liquid.

The application of the invention to transistor manufacture derives from the fact that junction transistors may consist of a crystal of germanium, or of other material such as silicon having like properties in this respect, with a controlled distribution of two particular significant impurities in it. Such a crystal is a semiconductor in which the distribution of the significant impurities produces diferent types of semiconduction in difierent parts of the crystal. (In the following description, germanium will be referred to as representing the various materials having these properties.) The impurities which have been called significant include one which provides an excess of electrons and thus produces the condition for excess semiconductions and another impurity which abstracts electrons and provides the conditions for deficit semiconduction. Such impurities are respectively called the donor (or donator) impurity and the acceptor impurity. The terms n-type and p-type are used to describe material capable of excess or deficit semiconduction,

respectively.

In a junction transistor, the distribution of the significant impurities is such that a thin layer across the crystal contains a preponderance of one type of impurity and is therefore either n-type or a p-type material, while the remainder of the crystal on each side of the layer con tains a preponderance of the other impurity and is the opposite kind of semiconductor material. Such a crystal may be referred to as a n-p-n semiconductor or a p-n-p semiconductor depending on which type of semiconduction characterizes the middle layer.

The object of this invention is to provide means for producing semiconductors having desired distributions of n-type and p-type conductivity regions, and in particular .toprovide semiconductors having the distribution "required 'for junction transistors and p-n junction recti- "tiers.

fill

2,824,030 Patented Feb. 1%, 1958 According to the invention, a crystal of a semiconducting material having regions of p-type and n-type semiconduction is produced by melting a oody of semiconductive material containing two significant impurities which have dilferent coefficients of distribution and which respectively produce p-type and n-type semiconductive material, applying a temperature gradient to a part of the length of the melt in the absence of convective mixing to cause linear solidification from one end to the other of the said part to give a predominance of concentration in the solid produced to each of the impurities in succession.

The need for such careful control of the conditions for solidification arises from the significant influence on the distribution of impurities in the solid exerted by difluson and convection in the melt. The concentration of impurity in the solid at any location is a function of the concentration of the liquid which during freezing was adjacent to the solid-liquid interface at that location, and since the impurities are rejected at the interface, the layer of liquid near the interface has a higher concentration of impurities than the main body of liquid. Therefore both diffusion and convection, by tending to remove this layer of high impurity concentration and spread it uniformly throughout the liquid phase, have a significant influence on the distribution of impurities in the solid.

When, according to the invention, a crystal is to be formed from a melt containing two solutes in low concentration, these solutes can be regarded as behaving independently for ranges of concentration of such as that to which the invention applies. The impurities and their concentrations are so chosen that when one of the impurities is included at a greater concentration in the melt than the other, this impurity which is predominant in the liquid is predominant in only a part of the solid formed, while the other impurity is predominant in another part. Since one of the impurities gives p-type semiconduction and the other gives n-type semiconduction, such a change in the relative concentrations of the two impurities in the solid will result in a change of conductivity type, with the result that a p-n junction is produced. The junction is produced during what will be referred to as transient phases of solidification; there are three such transient phases of solidification.

The three transient phases which can be taken advantage of to produce such junctions may be called the initial transient, the change in rate transient, and the terminal transient. The initial transient is the building up of the concentration of impurities in the solid during the first part of the solidification, producing a transient rise in the concentration of each impurity which continues until a steady state value C is reached in the solid. The change in rate transient is produced by an abrupt change in the rate of solidification, while the terminal transient represents the transient rise in concentration which is caused when the advancing solid-liquid interface, carry ing ahead of it the layer of liquid at the interface having the steady state impurity distribution C approaches the end of the melt at constant speed.

The object of the invention is attained by the invention as disclosed in the following description and accompanying drawings.

In the drawings-in which each reference character denotes the same part in all the views Fig. 1 is a diagrammatic representation of the process of solidification of a germanium crystal.

Fig. 2 shows graphically the calculated distribution of the significant impurities gallium and antimony in a germanium crystal for the method of preparation according to the invention using the initial transient. A

Fig. 3 shows graphically the calculated difference in concentration across a p-n junction of the significant impurities gallium and antimony in a germanium crystal for an initial transient.

Fig. 4 shows graphically the calculated distribution of the impurities resulting from the repetition, after mixing, of the initial transient phase.

Fig. is a diagrammatic view of apparatus for the production of single crystals of germanium containing p-n junctions suitable for use in rectifier or transistor manufacture.

The theoretical considerations on which the invention is based will now be discussed in order to explain the application of the transients to the production of crystals according to the invention. These theoretical considerations are based on the absence of convective mixing, and also on the further assumption that (1) Diffusion in the solid is negligible.

(2) The distribution coefficient of each impurity in the presence of the other (defined as the ratio of the concentration of the impurity in the solid being formed to the concentration in the liquid from which it forms), is a constant, and

(3) The interface separating the solid being formed and the liquid is plane and perpendicular to the axis of the specimen, as indicated diagrammatically in Fig. 1.

The rate of diffusion of impurity in the liquid is, as already mentioned, specifically taken into account.

Since the concentration of each of the significant impurities appearing in the solid is related by the constant factor k to that in the liquid adjacent to the interface, the distribution in the solid formed may be obtained by calculating the concentration in the liquid at the interface for all positions occupied by the interface during freezing. It has been assumed that the only factor causing movement of solute in the liquid is diffusion and, therefore, the equation which must be solved is the one-dimensional diffusion equation.

With regard to a fixed co-ordinate system, this equation may be written as:

However, the solute distribution relative to the moving solid-liquid interface is what is sought, and to obtain this it is convenient to use a co-ordinate system which moves with the interface. Transformed into this co-ordinate system, Equation 1 becomes:

where:

R=rate of movement of the interface, assumed to be constant x'=distance measured from the interface into the liquid For the case of a distribution coeflicient which is less than unity, the solid that forms will be purer than the liquid from which it freezes, which means that rejection of solute into the liquid will occur at the solid-liquid interface. Owing to the finite rate of diffusion of impurity in the liquid, a layer of high impurity concentration will build up adjacent to the advancing interface. As the concentration rises in the liquid at the interface, it must also rise in the solid being produced, since the two concentrations are related by the distribution coefiicient k. If the original melt is of uniform concen tration C thenthe layer of high impurity concentration must build up until the concentration appearing inlthe solid is also C the concentration in the liquid. at the interface will then be C /k. When this condition has been attained, the layer of high concentration will have reached a steady state, and no further variation with time will take place as long as the conditions of growth are not altered. Since time has been eliminated as a variable, the steady state distribution of impurity in the liquid may be found from a solution of subject to the boundary conditions:

01,: 0 at it) C =C /k at x=0 The solution may be written:

where q=(1-k).

THE INITIAL TRANSIENT In order to calculate the way in which the region of high solute concentration builds up adjacent to the interface during the first part of the solidification and the resulting transient rise of concentration to the steady state value, C in the solid produced, it is necessary to solve the time-dependent Equation 2. The solution will be carried out for the case of a semi-infinite sample in which solidification is initiated at one end at time t =0, and proceeds at constant speed R.

Equation 2 must be solved subject to the following boundary conditions:

C =C at as for all values OH (5) CL=O at 25:0 for all values of at greater than 0 (6) dCL Equation 9 may now be solved for 6, and by operation on 6 with the inverse Laplace transformation, the time dependent solution, C (x',t) is restored. The solution gives the concentration C throughout the liquid at. any time t.

The impurity concentration appearing in the solid is equal to k times the concentration in the liquid at x=0 at any time. Therefore, the impurity distribution resulting fromsolidification ofthe initial transient region. may be obtained from C (x',t) by putting x=0 multiplying by k, and introducing the coordinate x=Rt, which rep- APPLICATION OF THE INITIAL TRANSIENT As already mentioned, the invention utilizes the effect of the transients on the distribution in the crystal of two suitable impurities. In carrying out the invention, it is preferable to use the initial transient and this aspect of the invention will be explained in detail as being representative of it. In utilizing the initial transient, the choice of the significant impurities and their concentration depends upon the theoretical considerations set out below.

According to the invention, one of the chosen impurities, which may be called impurity 1 has a distribution coeflicient k which is greater than the distribution coefiicient k of impurity 2. Furthermore, the impurities must be so chosen that while the initial concentration C (1) of impurity 1 in the melt may be less than the corresponding initial concentration C (2) of impurity 2., nevertheless the initial concentration, expressed by the product lqC l), of impurity 1 in the crystal formed will be greater than the corresponding initial concentration k C (2) of impurity 2 in the crystal. Since (3 (1) and (3 (2) are the final equilibrium concentrations of the impurities in the crystal, this means that in at least the first part of the initial transient region of the crystal, impuritiy 1 will have the greater concentration, while in the remainder of the specimen, impurity 2 will be predominant.

Suitable significant impurities for germanium crystals made according to the invention, which produce p-type and n-type germanium respectively, are gallium and antimony. The coeflicients of distribution k and k of these elements have been given the values 0.12 and 0.007 respectively by W. G. Pfann, Journal of Metals, 4 (1952), 861. If the relative concentrations C (Ga) and C (Sb) of gallium and antimony in the melt are suitably regulated, C (Ga) will be less than C (Sb) while the value of the product 0.12 C (Ga) will be greater than the value of 0.007 C (Sb), and the germanium crystal formed by solidifying the melt will be p-type in the first portion to solidify (Ga predominant) and n-type in the remainder (Sb predominant).

The actual distribution of the impurities may be calculated from Equation using the known values of k and k and this distribution may be shown in graphical form by plotting the concentrations of gallium and antimony against the variable R/Dx in Equation 10. This has been done in Fig. 2, using various starting concentra tion ratios of C (Ga) to C (S b) for the impurities, and expressing the concentrations of both impurities in terms of the starting concentration C (Sb) of antimony in the melt, which is taken as unity; the starting value or" C /C for antimony is therefore k or 0.007. A p-n junction is formed where the atomic concentrations of antimony and gallium are equal.

It is clear from the graph that both the thickness of the layer of p-type garmanium and the difference between the two concentration gradients at the junction depend upon the rate of solidification R and the starting concentration ratio. Therefore the rate of change of the difference in concentration, which will be referred to as the differential concentration gradient, may be controlled in the neighbourhood of the junction by suitable control of the rate of growth and the starting concentration ratio. By subtracting corresponding concentration values in Fig. 2, the ditference in the concentrations of the two impurities may be calculated. This diiference is plotted in Fig. 3; the slope of the curve at the p-n junction gives the differential concentration gradient across the junction.

Once the region of solidification has passed the junction, the solid being formed is n-type germanium. To produce another layer of p-type material, it is merely necessary to break up the solute distributions which have built up in the liquid adjacent the interface, which may be done by stopping the solidification and stirring the liquid.

When the solidification process is resumed, the first material to solidify will again be p-type germanium. In other words, the initial transient region of solidification will be repeated, and a p-layer will be formed between two n-layers, as shown graphically in Fig. 4. The p-n junction formed at the point where the growth of the crystal resumed is a step-junction, while the other junction previously formed as described above, will be a graded junction. The result is a crystal from which a transistor may be made, provided that the p-layer is thin enough for that purpose. a

As already pointed out, the thickness of the p-layer may be controlled by varying the speed of solidification and the starting concentration of the two impurities. A further means of varying the thickness of the p-layer is to choose different solutes. To produce the thinnest possible p-layer, it is necessary that the point at which equal atomic concentrations of the two solutes occurs should be as close as possible to the point at which solidification is resumed.

This will be accomplished where the rise of the concentration of impurity 2 (the impurity having the lower distribution coefiicient) from k C (2) is as rapid as possible and the concentration of impurity 1 from k C (l') decreases; such a negative change in concentration will be produced at an increasing rate for increasing values of k that are over unity.

Calculations derived from Equation 2 show that the maximum rate of rise of concentration of an impurity will occur for a distribution coeflicient of /2. Thus to obtain a p-layer which is as thin as possible, the impurity with the lower distribution coefficient k should have a distribution coefficient as close as possible /2; and the other impurity should have a distribution coeificient which is as high as possible. Then the layer which is produced by impurity 1) will be as thin as possible where the distribution coefficient of this impurity is greater than unity. To produce a thin player in germanium, a suitable n-type impurity is arsenic (k =0.04) and a suitable p-type impurity is boron (k l), based on the values of the coefiicients given by Burton et al. in Journal of Chemical Physics (1953), 21, p. 1992.

If the melt is long enough, the steps of the process may be repeated several times and the finished crystal may be cut at suitable locations to produce a number of crystals suitable for use in junction transistors.

lf p-n junction rectifiers rather than transistor crystals are desired, cuts may be nade at other locations to produce crystals having one p-type and one n-type layer only.

PROCEDURE The apparatus illustrated in Fig. 5 consists of a Wire wound furnace A having main furnace windings A and auxiliary furnace windings A which are arranged in such a way that the hottest point of the furnace will be. close to the end where the auxiliary winding A is located. Thus a continuous temperature gradient exists over a large part of the length of the furnace. A copper cooling jacket A is also incorporated to provide additional control over the temperature gradient.

A silica or vycor tube B extends through the furnace Within the windings and extends beyond the furnace at each end, and a boat C of pile-purity graphite, silica, or other suitable refractory material is placed in the tube; the boat can be moved by means of a rod H ego tending through one end of the tube. (The rod should be made of a material such as inconel or graphite which will withstand the high temperatures in the furnace.) The charge of germanium or like material E is placed in the boat C with a single-crystal seed of germanium D at one end. A stirrer F extends into one end of the boat and is operated from outside the tube by means of a handle extending through a gas-tight seal in one of the caps G at the end of the tube. Small tubes for introducing and leading oif an inert atmosphere extend through the caps G, which may be water-cooled.

In operation, a suitable charge of germanium, having the proper concentration of impurities, is placed in the boat and a small flow of argon gas is passed through the tube. The temperature of the furnace is raised until the charge of germanium is melted and becomes fused onto the seed crystal, part of which is also melted in the process. Care must be taken to avoid melting more than a small portion of the seed crystal. The temperature is maintained at a constant value while the melted charge is stirred to ensure a uniform distribution of the impurities in it. Then the temperature of the furnace is slowly lowered to cause solidification of the melt at the desired rate. When a sufficient length of the melt has been solidified, to establish a region of one type of conductivity and it is desired to form a step-junction, the temperature of the furnace may be raised slightly to stop solidification, preferably re-melting a small portion of the length already solidified in order to ensure that solidification has stopped. The melt is again stirred to restore a uniform distribution of impurity and slidification is continued as before. The cycle of stirring and solidification is repeated until the whole sample has solidified.

The production of a layer of p-type germanium between two regions of n-type germanium involves the following steps:

(1) An ingot of germanium is prepared containing suitable quantities of suitable impurities, such as antiroomy and gallium.

(2) The ingot is placed in a suitable boat, melted and fused to a seed crystal. Melting is then carried out in an inert atmosphere such as argon.

(3) solidification of the melt is allowed to proceed from one end without stirring, in order to obtain a layer of high impurity concentration in the liquid adjacent to the advancing solid-liquid interface. This will establish the conductivity type of the material freezing out.

(4) The solidification is stopped and the liquid is sturred to destroy the high-concentration layer adjacent to the interface.

(5) The solidification is resumed, thereby producing a layer of p-type germanium in the manner described in detail in the previous report.

(6) Steps 4 and 5 are repeated at intervals until the entire sample is solidified.

Example I Preparation of ing0t.-The impurities used in this experiment were antimony, which produces n-type germanium, and gallium, which gives p-type germanium. The concentrations of these impurities required in the ingot to be used for the experiment were very low, so

that it was impractical to attempt to add them directly.

' horizontally on two retort stands.

Sb master alloy: 3.2 l0 Sb atoms per gram. Ga master alloy: 5.4)(10 Ga atoms per gram.

Auxiliary experiments had shown that some contamination of pure germanium occurred when it was melted in a silica or graphite container. This appeared to be unavoidable, and in order to minimize the effect, antimony and gallium concentrations were used which were much higher than the residual impurity concentrations. Nominal concentrations chosen were:

Gallium concentration in melt =C (Ga) 1.2x 10 atoms/cm. :27 (2) X 10* atomic percent. Antimony concentration in melt :C (Sb) =5.9 10 atoms/cm. 1.3 (4) X 10* atomic percent.

To obtain these concentrations, the following quantities were weighed:

Ge (resistivity greater than 35 ohm-cm.) 12.222 Ga master alloy 0.839 Sb master alloy -a 6.942

On the basis of these Weighed quantities, the resulting concentrations in the melt will be:

C Ga) =2.7( 3) l0 atomic percent. C (Sb) =1.3(4) 10 atomic percent.

This gives a concentration ratio Seed crystal.The seed used in this experiment was not a single crystal but rather a polycrystalline bar made by melting pure germanium (25 ohm-cm. or higher resistivity) in the graphite boat to form a bar of suitable dimensions. It was not considered necessary to use a single crystal seed for this experiment since the impurity distribution to be obtained would not depend to any appreciable extent upon the state of polycrystailinity or" the seed. It was known from other experiments that a single crystal seed could be successfully used if desired.

The copper cooling jacket was soldered to one end of the seed using pure lead as soldering material. Water cooling applied by means of this jacket served two purposes. First, it allowed the estabiishment of a steep temperature gradient along the sample so that solidification could be started rapidly when desired. Secondly, it improved the linearity of heat flow along the sample, thereby providing an approximately plane solid-liquid interface during solidification.

Pr0cedure.-The seed, with its cooling jacket, and the charge of doped germanium, were placed in the graphite boat. This assembly was then slipped into a Vycor (temperature-resistant glass) tube which was mounted The Vycor tube was sealed at each end with a rubber stopper through which passed the tubes carrying cooling water to the seed cooling jacket and argon gas to the interior of the tube.

The furnace was a small electrical resistance furnace placed around the Vycor tube at the position of the graphite boat and mounted on wheels and tracks so that it could be moved parallel to the tube. The furnace temperature was controlled by a Minneapolis-Honeywell pro gramming controller which allowed the temperature to be maintained at any desired value or to be lowered at any desired rate. The temperature was raised to approximately 1-050 C. (measured at the outside of the Vycor tube above the germanium charge). This temperature was sufficient to melt the germanium charge and fuse it onto the seed. Approximately 4 inch of the seed was melted to assure a good joint. The volume of seed melted was small compared to the volume of the doped charge so that the gallium and antimony concentrations would be changed by only a small amount. Since the seed was pure germanium, the concentration ratio C (Ga) o would not be changed.

As soon as the charge was fused onto the seed, the furnace was moved about 2 inches away from the seed and, at the same time, the furnace temperature was reduced at a rate of 400 C./hour by means of the programming controller. These two factors caused solidification to occur at a speed of about 3 mm. per minute. When about 1.5 cm. of the sample had been solidified, to extend the seed for future use and to establish a region of n-type germanium, the furnace was moved back about 1" and the temperature was raised enough to stop solidification and to re-melt about 0.5 cm. of the solid. The liquid was then agitated to ensure that its composition became uniform, and solidification was resumed by again moving the furnace away 1 to 2 inches and lowering its temperature at a rate of 400 C. per hour.

This cycle of stopping solidification, stirring, and then resuming solidification was carried out twice more during freezing of the sample before the furnace power supply was then shut off and the sample allowed to cool to room temperature in the furnace.

Examination of the sample-Using a diamond saw, the sample was cut perpendicular to its axis to give three sections, each about 1 cm. in length and each containing one of the three regions in which the cycleof stirring and solidification had been carried out. Each of these sections was in turn cut in a plane parallel to the axis and perpendicular to the bottom of the sample to produce two internal surfaces on which the expected player should appear. The cut surfaces were then etched in asolution consisting of 3 parts glacial acetic acid, 5 parts concentrated nitric acid, and 3 parts hydrofluoric acid (known as CP4 A etching solution). This etching reagent revealed as a fine line on the cut surface of each section the step junction produced by stopping solidification, stirring and resuming solidification and representing an abrupt change of conductivity type, as discussed in a previous report. The location of the graded junction, where there is a gradual change of conductivity type, the other side of the p-layer, is not revealed by the etching.

In order to measure the thickness of the p-type layer of germanium, it was necessary to determine the conductivity type as a function of distance along the sample 'in the appropriate region. This was accomplished by making use of the fact that the thermal emf. generated by a copper to n-type germanium contact is of opposite sign to that generated by a copper to p-type germanium contact. A hot probe was constructed by winding a small heating .coil on a sharply-pointed piece of copper wire. When this hot probe was placed in contact with the germanium sample, a hot contact was created, thereby generating a thermal emf. which could be detected by :means'of :amillivoltmeter to give a determination of con ductivity type. The hot probe was mounted in a micromanipulator which allowed it to be moved along the "sample bysuccessive distances as small as 0.01 mm.

The results of measurements of the p-layer thicknesses produced in this experiment are given in Table I below. The :p-layers are numbered in order from the seed end of thespecimen. Measurements were made on each of the cut sections near the top, centre and bottom of the sample.

TABLE I Speed of p-layer Position Solidifica- Thicktion, ness, mm. mm./ml.n.

Example II In this example the acceptor and donor impurities used to produce p-type and n-type germanium respectively were boron and arsenic. Where applicable, the general procedure used was essentially the same as in Example I and will not be discussed in detail.

Preparation of arsenic master alloy-Since arsenic sublimes at about 615 C. under atmospheric pressure, and only melts at about 814 C. under a pressure of 36 atmospheres, a Ge -As alloy cannot be made by melting the two metals together under ordinary conditions. Accordingly, it was decided to make this alloy by bringing together arsenic vapour and liquid germanium. The arsenic used was first purified by subliming a quantity of it twice. 10 mg. of the purified arsenic and 10 gm. of zone-refined Ge were placed in a silica tube which was then evacuated and sealed oif. The sealed tube was heated in a furnace until the germanium was melted. The charge was kept molten for about 1 hour and then allowed to freeze and cool to room temperature. No signs of arsenic deposited on the silica tube could be seen.

Preparation of boron master a'lloy.-The boron used was amorphous boron of about purity. This was alloyed with zone-refined germanium by the following procedure: 7 mg. of boron and 10 gm. of germanium were placed in a silica tube which was then connected to a vacuum system employing an oil dilfusion pump. The tube was evacuated and then heated to about 1000 C. in a furnace to melt the germanium. The germanium was .kept molten for about 24 hours and pumping was continued throughout this period.

The silica tube was then sealed otf under vacuum. Since there are some evidence of undissolved boron, the material was then heated to a higher temperature by induction heating in the sealed tube. This technique produced violent stirring of the liquid. A temperature of ll50'to 1175 C. was reached and maintained for 1 hour. The sample was then allowed to freeze and cool to room temperature.

Determination of impurity concentrations in master all0ys.-Since it was impossible to know how much of the arsenic and boron was actually dissolved by the germanium to form each master alloy, it was necessary to obtain some idea of the concentrations in these alloys. This was done by measuring the Hall constant of each master alloy. (The results of such measurements must of course be regarded as only a very approximate estimate of the impurity concentrations, particularly in the case of the boron alloy. The results will depend upon all of the impurities present in the material, and since the boron used was only 85% pure other impurities were undoubtedly introduced. While these impurities could not be the determining ones during the experiment, they could make it impossible to obtain more than a rough estimate of the boron content of the master alloy.)

The impurity contents of the master alloys, as measured by the Hall effect method, were:

Boron master alloy: 1.8)(10 atoms/cm. Arsenic master alloy: 2.4.x 10 atoms/cm.

Preparation of charge-Io avoid efiects of contamination from the boat in which the sample was to be solidified or other possible causes, a minimum impurity content of about l /cm. was chosen. To achieve this, approximate concentrations in the starting melt were required to be:

C (As)=6.25 As atoms/cm. and C (B) 1.25 x 10 B atoms/cm.

Two experimental runs were carried out. The amounts of material used in each were as follows:

Run #1:

Ge( ohm-cm. or higher) 14.352 As master alloy 0.0378 B master alloy 0.0101

Total charge"; 14.3999

According to estimated concentrations in the master alloys, this should give concentrations in the initial melt of As=6.3 (0) X l0 /cm. B=1.2(7) 10 /0111.

Estimated concentrations will be:

Procedure.For each run the charge, along with a water-cooled pure germanium seed crystal, was placed in the graphite boat which was then inserted in the vycor tube of the furnace assembly. Gas and water connections were made, the vycor tube was flushed with argon gas, and the cooling water and furnace power were turned on. The furnace temperature was raised to about 1050 to 1075" C. to melt the charge and fuse it to the seed crystal.

Then growth of the specimen was carried out in accordance with the cycle required to produce the desired players. In each case, growth was promoted by pulling the furnace off by l to 2 inches and, at the same time, commencing a lowering of the furnace temperature at a rate of 400 C. per hour by means of the programming controller. In order to assure adequate mixing, a portion of the sample was re-melted each time mixing was carried out.

Examination of samples.---Using a diamond Saw, a vertical longitudinal section was cut through each region in which a p-layer was expected to be located. Each section was examined by etching to reveal the location of the step junction and by means of the hot probe to determine conductivity type. Measurements of the thickness of some of the p-layers were carried out using the hot probe mounted in a micromanipulator.

Results.--Characteristic results are tabulated below:

It must be noted that the roughly estimated values of the boron concentration in run #2 appeared to be greater than that of the arsenic concentration. Under these conditions, according to theory, the sample would be p-type throughout. However, the consistency in the results of run #2, shows that the actual concentrations were within the theoretical requirements.

PRODUCTION OF REGIONS OF UNIFORM COMPOSITION A supplementary aspect of the invention is the production of crystal zones which are of uniform composition. To do this it is necessary to eliminate all transient regions of solidification within such zones. To obtain such zones of maximum length, transient regions are minimized in length by growth at high speed, and the speed of solidification must be maintained constant throughout the solidification to eliminate the transient changes in composition which accompany changes in speed.

The foregoing description sets forth the best mode con templated by the inventors of carrying out their invention, but the following claims are intended to cover all useful changes and modifications of the said mode which are within the scope of the invention.

What we claim as our invention is:

1. A method for preparing semiconductive material having regions of p-type and n-type semiconduction, comprising melting a body of semiconductive material containing substantially uniform solute concentrations C (l) and C (2) of two significant impurities l and 2 that have different coefficients of distribution k and k respectively in the said material, one of which impurities produces p-type and the other n-type semiconductive material, the initial concentration C (l) of one impurity being less than the corresponding concentration C (2) of the other impurity and the product C (l) being greater than the product k C (2), applying a temperature gradient to a part of the length of the melt in the absence of convective mixing to cause linear solidification from one end to the other of the said part and to establish and maintain a liquid-solid interface having an adjacent layer of molten solute-rich material having concentrations C (1) and C (2) respectively of the said impurities whereby the concentrations k C (l) and k C (2) respectively in the solid produced are varied until steady-state concentrations C (1) and (3 (2) re.- spectively of the said impurities are established in the solid, whereby the concentration of impurity 1 inthe solidified material will first be greater than, thenless than, the concentration of impurity 2, mixing the remaining, liquid, part of the melt to destroy the layer of solute-rich material at the interface and produce a uniformity of solute concentration in the liquid, and applying a temperature gradient to the said remaining part of the melt to cause linear solidification to proceed from the solid-liquid interface in the absence of convective mixing.

2. A method for preparing semiconductive material as claimed in claim 1 in which the melt is of generally uniform cross-section.

3. A method for preparing semiconductive material as claimed in claim 1 in which the semiconductive ma terial is germanium and the impurities having coefiicients of distribution k and k are gallium and antimony respectively.

4. A method for preparing semiconductive material having a relatively thin layer of p-type conduction between two regions of n-type conduction, comprising melting a body of germanium containing a relatively small uniform concentration of antimony and a smaller uniform concentration of gallium, the product of the distribution coefficient and the concentration of gallium being greater than the product of the distribution coefiicient and the concentration of anitmony, applying a temperature gradient to part of the length of the melt in the absence of convective mixing to cause linear solidification from one end to the other of the said part and to establish and maintain a liquid solid interface having an adjacent layer of molten solute-rich material until the concentrations of antimony and gallium in the solid are respectively substantially equal to the concentrations of gallium and antimony in the body of germanium prior to melting, mixing the remaining liquid to destroy the layer of solute-rich material at the interface and produce a substantially uniform concentration of gallium and antimony throughout the liquid, and repeating the solidification from the solid-liquid interface in the absence of convective mixing. I

5. A method for preparing semiconductive material as 2,567,970 Scafi et a1 Sept. 18, 1951 2,615,060 Marinace et al. Oct. 21, 1952 2,711,379 Rothstein June 21, 1955 2,739,088 Pfann Mar. 20, 1956 FOREIGN PATENTS 1,065,523 France Jan. 13, 1954

Claims (1)

1. A METHOD FOR PREPARING SEMICONDUCTIVE MATERIAL HAVING REGIONS OF P-TYPE AND N-TYPE SEMICONDUCTION COMPRISING MELTING A BODY OF SEMICONDUCTIVE MATERIAL CONTAINING SUBSTANTIALLY UNIFORM SOLUTE CONCENTRATIONS C0(1) AND C0(2) OF TWO SIGNIFICANT IMPURITIES 1 AND 2 THAT HAVE DIFFERENT COEFFICIENTS OF DISTRIBUTION K1 AND K2 RESPECTIVELY IN THE SAID MATERIAL, ONE OF WHICH IMPURITIES PRODUCES P-TYPE AND THE OTHER N-TYPE SEMICONDUCTIVE MATERIAL, THE INITIAL CONCENTRATION C0(1) AND ONE IMPURITY BEING LESS THAN THE CORRESPONDING CONCENTRATION C0(2) OF THE OTHER IMPURITY AND THE PRODUCT K1C0(1) BEING GREATER THAN THE PRODUCT K2C0(2), APPLYING A TEMPERATURE GRADIENT TO A PART OF THE LENGTH OF THE MELT IN THE ABSENCE OF CONVECTIVE MIXING TO CAUSE LINEAR SOLIDIFICATION FROM ONE END TO THE OTHER OF THE SAID PART AND TO ESTABLISH AND MAINTAIN A LIQUID-SOLID INTERFACE HAVING AN ADJACENT LAYER OF MOLTEN SOLUTE-RICH MATERIAL HAVING CONCENTRATIONS CL(1) AND CL(2) RESPECTIVELY OF THE SAID IMPURITIES WHEREBY THE CONCENTRATIONS K1CL(1) AND K2CL(2) RESPECTIVELY IN THE SOLID PRODUCED ARE VARIED UNTIL STEADY-STATE CONCENTRATIONS C0(1) AND C0(2) RESPECTIVELY OF THE SAID IMPURITIES ARE ESTABLISHED IN THE SOLID, WHEREBY THE CONCENTRATION OF IMPURITY 1 IN THE SOLIDIFIED MATERIAL WILL FIRST BE GREATER THAN, THEN LESS THAN, THE CONCENTRATION OF IMPURITY 2, MIXING THE REMAINING, LIQUID, PART OF THE MELT TO DESTROY THE LAYER OF SOLUTE-RICH MATERIAL AT THE INTERFACE AND PRODUCE A UNIFORMITY OF SOLUTE CONCENTRATION IN THE LIQUID, AND APPLYING A TEMPERATURE GRADIENT TO THE SAID REMAINING PART OF THE MELT TO CAUSE LINEAR SOLIDIFICATION TO PROCEED FROM THE SOLID-LIQUID INTERFACE IN THE ABSENCE OF CONVECTIVE MIXING.

Images