US3167512A

US3167512A - Method of controlling the distribution of doping substance in crucible-free zone-melting operations

Info

Publication number: US3167512A
Application number: US814634A
Authority: US
Inventors: Ziegler Gunther
Original assignee: Sicmens & Halske AG
Current assignee: Sicmens & Halske AG
Priority date: 1958-05-21
Filing date: 1959-05-20
Publication date: 1965-01-26
Anticipated expiration: 1982-01-26
Also published as: CH369830A; FR1226810A; DE1164680B; GB902016A

Description

Jan. 26, 1965 G. ZIEGLER 3,167,512

METHOD OF CONTROLLING THE DISTRIBUTION OF DOPING SUBSTANCE IN CRUCIBLE-FREE ZONE-MELTING OPERATIONS Filed May 20, 1959 3,167,512 METHGD F CGNTROLLING THE DISTRIBUTION OF DOPING SUBSTANCE IN CRUClBLE-FREE ZQNE-MELTING QPERATIONS Giinther Ziegier, Munich, Germany, assignor to Siemens & Halske Aktiengesellschaft, Munich, Germany, a corporation of Germany Filed May 20, 1959, Ser. No. 814,634 Claims priority, application Germany May 21, 1958 8 Claims. (Cl. 252-623) This invention relates to an improvement in the processing of high-purity semiconducting materials. It particularly concerns processes in which a limited zone of a rod-shaped body of the material is melted and the melted zone caused to travel along the body.

The main object of the invention is to devise a reliable method for controlling the distribution of lattice-defection impurities, usually called doping substance, in the zonemelted body in order to obtain a desired characteristic, or type, of extrinsic conductance.

The following assumptions are often made in respect to the operation of the known zone-melting method:

(a) The diilfusion of the foreign substances in the solidified rod portion is negligible,

(b) The concentration C in the melting zone is uniform,

(0) The distribution coeflicient k is constant, and

(d) No reaction of the semiconductor material takes place with the crucible material and the atmosphere.

In many cases these assumptions are only partially met. This applies particularly to the assumption (b) of a uniform concentration in the melt and (d) the absence of reaction between crucible material and atmosphere. While an enrichment in concentration at the solidifying boundary face can be taken into account by introducing into the calculation an effective distribution coefiicient dependent upon the processing parameters, such a simple form of correction is not possible relative to reaction with the crucible material or atmosphere. lust these influences however are often the source of considerable disturbance.

The influence of the crucible material can be avoided by employing the crucible-free zone-melting method. The contamination of the semiconductor material by the gaseous atmosphere can be eliminated by operating in high vacuum. However, there then occurs an additional evaporation of the impurity substances which superimposes it self upon the segregation of material from the gas onto the semiconductor body, and thus fundamentally changes the shape of the concentration curves.

In the mathematical calculation of the evaporation effect, it is known to supplement the differential equation relating to the distribution of the lattice-defection points within a rod-shaped semiconductor body being zonemelted in a crucible, by adding to the equation an evaporation term. The equations of practical importance are presented below. The variation of the concentration C versus time, caused in the melting zone by evaporation, is taken into account by the linear equation This equation is tantamount to the assumption that the partial pressure of the dissolved substance above the solunited States Patent 3,.ib?,5i2 Patented Jan. 26., 1965 u k H+ZCtZC0 wherein the following relation was set to apply:

k 0: u F TZ (3) In Equation 2, the term C, denotes the concentration in the rigid, solidified end of the rod, and L the length of the melting zone; in Equation 3, v denotes the velocity with which the melting zone is passed along the rod.

The solution, when entering the corresponding limit conditions C =C for x=0, appears for a constant value of C and a constant speed v, as follows:

It follows from Equation 3 that:

a-L u k T The magnitude k is designated in the specification as the distribution coefficient which, as is well known, is defined as the ratio of the concentration of the impurity in the solid material to the purity contained in the melt. The values characteristic of the semiconductor material being employed, and of the impurities to be segregated, are known from published papers concerning zone melting and consequently are available to those skilled in the art. The zone length L and the zone-pulling speed v are known and observable parameters of the zone-melting method.

To facilitate visualizing the shape of the resulting curve, the graph in FIG. 1 indicates a number of concentration curves for the same distribution coefiicient k but for different values of the evaporation factor u. The starting point of the curves C C -k is independent of 0:. Then, a stationary value adjusts itself exponentially. As is apparent from FIG. 1, an originally homogeneous distribution of concentration, i.e. C =constant, can be kept homogeneous during the zone-melting (k-C if, by suitable choice of the zone length and zone-pulling speed, the value a is made equal to unity (u=1).

It is thus possible, by suitable choice between zone length and pulling speed, to eliminate the gradient of doping distribution along a semiconductor rod, which gradient results during zone-melting for producing a monocrystal, due to the fact that all impurities having the distribution factor k, different from the value 1, are transported toward the two ends of the rod.

According to the invention, therefore, a rod-shaped semiconductor body of extreme purity with a desired curve of the lattice-defection conductance and of the desired conductance type is produced by first excessively doping the high-purity conductor rod along a large portion of its length so that its conductance is greater than corresponds to the one ultimately desired, and thereafter subjecting the excessively doped rod to crucible-free zonemelting, preferably in a continuously evacuated vessel with small partial pressures of the doping substance. That is, the partial pressures of the doping substances con tained in the rod are below the respective vapor pressures of these substances at the temperatures of the melting zone. Furthermore, during such zone-melting, the percentile evaporation quantity of the lattice-defection points per unit of rod length is varied for the purpose of obtaining a desired, preferably constant, distribution of the lattice-defection points along the rod. This can be done, for example, by varying the speed and/or the length and/or the temperature of the melting zone during a passage of the melting zone through the body.

At the beginning of the melting-zone passage there occurs a decline in the defection concentration in the melting zone adjacent to the crystal germ. In order to prevent this to a large extent, and hence to obtain the desired lattice-defection distribution along the rod with a smallest possible number of zone-melting passes, it is preferable to keep the molten zone at first in stationary condition.

In addition, the crucible-free zone pulling, when employed in conjunction with the method according to the invention, aifords considerable advantages in comparison with melting in a crucible. As mentioned, the detrimental effect of the crucible material is avoided. It is of further importance for the vaporization that, with this method, the free surface is large relative to the volume of the melting zone, and that simultaneously a uniform evaporation and thus also an improved uniformity of concentration over the cross-section of .the semiconductor rod is secured. The zone-pulling in vacuum has the further advantage that impurities, such as oxygen, that are detrimental to the formation of the mono-crystal will likewise evaporate at the rod surface so that only a few zone passes are required to obtain the desired defection-point distribution as well as to produce the mono-crystal, which renders the method according to the invention especially economical. Furthermore, the starting concentration of the detection point prior to zone-melting that is required for the desired ultimate concentration can morereadily be calculated in advance, which likewise permits the reduction of the number of the zone-pulling passes to a minimum.

The new method is further explained as follows. The starting material can be obtained by a known method, such as by decomposing a gaseous compound of the semi conductor material in an electric are, or by means of the method described in French Patent 1,125,207, of October 26, 1956, namely by decomposing a gaseous semi-- conductor compound upon the surface of a heated carrier. For the latter method, note also Schweickert et a1. application Serial No. 736,387, filed May 19, 1958, now Patent No. 3,030,189. In both cases the over-doping required by the present invention is obtained by the simultaneous segregation of doping substances from gaseous compounds of the doping substances that are mixed with the gaseous semiconductor compounds. According to the invention, care is taken that the over-doping along the length of the rod is essentially constant, i.e. that departures amount to at most about 1*:10% from the average value along the entire rod.

Subsequently, the desired, preferably constant, distribu- 4 tion of defection points is obtained by applying a cruciblefree zone-melting, preferably in vacuum or, if desired, in an inert gas such as hydrogen, at normal atmospheric pressure. The actual length of the molten zone and the.

pulling speed are so adapted to each other that local accumulation of impurity substances, responsible for the otherwise observable doping gradient, is avoided and particularly that the ultimate doping is essentially constant over the length of the rod, any departures being within about i 10% from the median value along the entire rod.

When performing the zone-melting in a protective gas atmosphere, for example hydrogen, the amount of evaporation is reduced. This is understandable because in the presence of a protective gas a gas layer having the partial pressure of the evaporating substance will rapidly build up over the molten zone, while in vacuum the evaporating particles can fly off unimpededly. This gaseous layer is essentially affected by convection and diffusion. Since the amount of convection depends upon the design of the equipment, a similar dependence of the evaporation coefiicient upon the apparatus design is also to be expected when performing the zone-melting in a protective gas.

By selecting a ditferent combination of evaporation and segregation, any desired concentration distribution of the doping substance along the rod can be obtained if care is taken that the percentile reduction of the conductance, particularly by variation of the zone-pulling speed along the length of the body, is kept constant, preferably during each pulling pass but at least during one pass.

It is thus possible, for example, to obtain steps in the concentration curve by heating one zone of the rod to the melting temperature for a longer interval of time than the adjacent zone. Furthermore, a concentration gradient occurring only locally can be levelled by applying a longer evaporation interval.

In accordance with a specific example of the method, a silicon rod is at first excessively doped for n-type conductance and is subsequently converted to a desired, particularly constant, course of the n-type conductance by the above-described crucible-free zone-melting operation. For example, the rod was over-doped with phosphorus. The crucible-free zone-melting was performed in high vacuum (10? mm. Hg) by inductive heating of the melting zone, while one end of the rod, namely the one that solidifies last when performing the method, was being turned about its axis at approximately revolutions per minute, for the purpose of promoting the thorough mixing of the melt and thereby preventing the formation of a depletion zone at the surface. in order to prevent a reduction'in lattice-defection concentration at the beginning of the zone-pulling operation, the molten zone Was at first kept at rest for approximately 1.5 minutes, whereafter the forward feeding motion of the molten zone was commenced. This zone-melting is carried out in an apparatus the same as or similar to those described in the applications of R. Emeis, Serial No. 727,610, filed April 10, 1958, now Patent 2,972,525, and Serial No. 409,610, filed February 11, 1954, now Patent 3,030,194.

A reverse doping by vaporization of the phosphorus molecules from parts of the apparatus back to the molten zone is prevented by the fact that the phosphorus molecules are in part built into, i.e. incorporated into, the growing coating of silicon that is directly vapor-deposited upon parts of the apparatus, and are partly sucked off by the dynamic vacuum being used. Furthermore, a reaction of the phosphorus molecules can take place with the traces of oxygen contained in the vacuum.

FIG. 2 illustrates a semiconductor body, for example of silicon, made according to the invention, for use in a rectifier or transistor. The semiconductor body cornprises two p-n junctions denoted by 2 and 3. The nconducting zone 1 of the monocrystalline silicon body consists of a circular wafer obtained by slicing a semiconductor rod made by the method according to the invention.-

The crystalline silicon remains unchanged within the wafer body 1, whereas the p-zones are formed by doping them with acceptor substance thus converting the originally n-type silicon into p-type silicon. The reverse doping can be performed, for example, by alloying acceptor substances 4 and 5 into the originally n-conducting silicon, so that the

recrystallization zones

41 and 51 form the p-zones. As shown in FIG. 2, it is preferable that the zone junction 3, which acts as emitter, have a smaller cross-section than the zone 2 which is to serve as collector.

All of the pertinent disclosure of the above-identified U.S. applications is incorporated herein by reference, and reservation is made of right to insert the pertinent subject matter from the drawings and description of said applications should it be necessary or expedient to do so.

I claim:

1. A method for producing a rod-shaped semiconductor body of high purity having a desired constant course of the lattice-defection conductance and of the conductance type over a substantial portion of the length of the body, comprising subjecting a semiconductor rod-shaped body which is overdoped over a substantial portion of its length to zone-melting, the floating zone-melting being a cruciblefree zone-melting, in which the rod is vertical and the floating molten zone is caused to be relatively displaced lengthwise of the rod, and being carried out in a chamber, continuously evacuating the chamber and thereby keeping the body under a partial vapor pressure of the doping sub stance which is below the vapor pressure of said substance from said body, so that evaporation of said substance from the body can take place, the percentile evaporating quantity of the lattice-defection points per unit length of the body being varied during the passage of the melting zone through the body to obtain a substantially constant defection distribution along the body, said varying being accomplished by varying at least one of the following: the speed of passage of the melting zone, the length of said zone, and the temperature of said zone, so that the value of is substantially one, k being the distribution coefficient, L the length of the melting zone, v the velocity with which the melting zone is passed along the rod, and a the evaporation constant, being the proportionality factor in the linear differential equation:

C being the concentration of impurities in the melting zone.

2. The method according to claim 1, the overdoping along the length of the rod having departures from constancy which range from negligible to at most about il0% from the average value over the entire rod.

3. The method according to claim 1, the ultimate doping along the length of the rod having departures from constancy which range from negligible to at most about il0% from the average value over the entire length.

4-. A method according to claim 1, the percentile reduction in conductance being kept substantially constant by varying the zone-travelling speed over the length of the body during each passage.

5. The method of claim 1, the semiconductor being germanium.

6. A method for producing a rod-shaped silicon semiconductor body of high purity having a desired constant course of the lattice-defection conductance and of the conductance type over a substantial portion of the length of the body, comprising producing a semiconductor rodshaped body which is overdoped over its length by decomposing a mixture of a reducing gas, a gaseous silicon halogenide, and a vaporized compound of phosphorus that decomposes upon heating into phosphorus, the de composing being upon a silicon rod brought to decomposing temperature by electric current flowing therein, subjecting the body to zone-melting to form a monocrystal, the zone-melting being a crucible-free zone-melting and being carried out in a chamber which is evacuated con tinuously, the body being under a partial vapor pressure of the doping substance which is below the vapor pressure of said substance from said body, so that evaporation of said substance from the body can take place, the percentile evaporating quantity of the lattice-defection points per unit length of the body being varied during the passage of the melting zone through the body to obtain a substantially constant defection distribution along the body, said varying being accomplished by varying at least one of the following: the speed of passage of the melting zone, the length of said zone, and the temperature of said zone, so that the value of is substantially one, k being the distribution coeflicient, L the length of the melting zone, v the velocity with which the melting zone is passed along the rod, and a the evaporation constant, being the proportionality factor in the linear differential equation:

t --aCs C being the concentration of impurities in the melting zone.

7. The method of claim 6, the doping substance being phosphorus.

8. A method of doping a semiconductor silicon rod uniformly over its length for a given type of conductance, comprising overdoping the rod over a substantial part of its length with a doping substance conferring said type of conductance, distributing the doping substance substantially uniformly over the length of the rod by subjecting the rod to a crucible-free floating zone-melting operation in which the rod is vertical and the molten zone is caused to be relatively displaced along the rod length, the molten zone being maintained under high dynamic vacuum to promote vaporization and sucking off of the doping substance from the rod, the speed of said displacement, and the length and temperature of the molten zone being adjusted so as to obtain said uniform doping, so that the value of ozL 10 +7 is substantially one, k being the distribution coefficient, L the length of the melting zone, v the velocity with which the melting zone is passed along the rod, and a the evaporation constant, being the proportionality factor in the linear differential equation:

C being the concentration of impurities in the melting zone.

References Cited in the file of this patent UNITED STATES PATENTS 2,739,088 Pfann Mar. 20, 1956 2,792,317 Davis May 14, 1957 2,813,048 Pfann Nov. 12, 1957 2,849,343 Kroger Aug. 26, 1958 OTHER REFERENCES Keck, Horn, Soled, and MacDonald: The Review of Scientific Instruments, volume 25, No. 4, 331-334, April 1954.

Claims

1. A METHOD FOR PRODUCING A ROD-SHAPED SEMICONDUCTOR BODY OF HIGH PURITY HAVING A DESIRED CONSTANT COURSE OF THE LATTICE-DEFECTION CONDUCTANCE AND OF THE CONDUCTANCE TYPE OVER A SUBSTANTIAL PORTION OF THE LENGTH OF THE BODY, COMPRISING SUBJECTING A SEMICONDUCTOR ROD-SHAPED BODY WHICH IS OVERDOPED OVER A SUBSTANTIAL PORTION OF ITS LENGTH TO ZONE-MELTING, THE FLOATING ZONE-MELTING BEING A CRUCIBLEFREE ZONE-MELTING, IN WHICH THE ROD IS VERTICAL AND THE FLOATING MOLTEN ZONE IS CAUSED TO BE RELATIVELY DISPLACED LENGTHWISE OF THE ROD, AND BEING CARRIED OUT IN A CHAMBER CONTINUOUSLY EVACUATING THE CHAMBER AND THEREBY KEEPING THE BODY UNDER A PARTIAL VAPOR PRESSURE OF THE DOPING SUBSTANCE WHICH IS BELOW THE VAPOR PRESSURE OF SAID SUBSTANCE FROM SAID BODY, SO THE DOPING SUBSTANCE WHICH IS BELOW THE VAPOR PRESSURE OF SAID SUBSTANCE FROM SAID BODY, SO THAT EVAPORATION OF SAID SUBSTANCE FROM THE BODY CAN TAKE PLACE, THE PERCENTILE EVAPORATING QUANTITY OF THE LATTICE-DEFECTION POINTS PER UNIT LENGTH OF THE BODY BEING VARIED DURING THE PASSAGE OF THE MELTING ZONE THROUGH THE BODY TO OBTAIN A SUBSTANTIALLY CONSTANT DEFECTION DISTRIBUTION ALONG THE BODY, SAID VARYING BEING ACCOMPLISHED BY VARYING AT LEAST ONE OF THE FOLLOWING: THE SPEED OF PASSAGE OF THE MELTING ZONE, THE LENGTH OF SAID ZONE, AND THE TEMPERATURE OF SAID ZONE, SO THAT THE VALUE OF