JPH025716B2 - - Google Patents

Description

[Detailed description of the invention]

(Description of related applications) TRanthony
U.S. Patent Application No. 1, filed June 11, 1981, and assigned to the same assignee as the present application.
No. 272801 discloses anode and cathode contacts useful in general for moving metal-containing melts through the use of potential gradients, and particularly useful for avoiding the negative effects of various rectifying junctions. A method of formation is disclosed. The contents of the above application specifications are hereby incorporated by reference. FIELD OF THE INVENTION The present invention relates to a method of moving a metal-containing melt through a semiconductor material through the use of a potential gradient, thereby forming regions of recrystallized single crystal semiconductor material in the semiconductor material. . More particularly, the present invention relates to the formation of PN junctions by such methods. BACKGROUND OF THE INVENTION The method of transferring metal-containing melts in solids using temperature gradients, the temperature gradient zone melting method (TGAM method), was invented by Pfann and is described in, for example, U.S. Pat. No. 2,813,048.
It is stated in the specification of the issue. According to Pufmann's method, a sheet or wire of a suitable liquid material is moved through a semiconductor by means of a temperature gradient. As the liquid metal wire continues to move, impurity-doped liquid phase epitaxial material is left behind. Subsequently, several researchers discovered that planar liquid zones are unstable and that droplets from such zones can break off and become trapped in recrystallized semiconductors. For these reasons, Puhuan's TGZM
Materials processed by this method were not suitable for use in the manufacture of semiconductor devices. More recently, however, Anthony & Cline have begun commercial efforts to form PN junction structures in semiconductors, such as the isolation lattices described in U.S. Pat. We have developed a TGZM method that can be implemented in a practical manner. newly developed
For details of the TGZM method, apparatus, and PN junction structures formed thereby, see, for example, US Pat. Briefly, the commercially modified TGZM process involves placing a semiconductor wafer between an array of air-cooled, high-temperature quartz lamps and a water-cooled heat sink.
Infrared radiation from the lamp is absorbed by the wafer surface facing the lamp and converted into heat. After passing through the wafer, this heat is radiated from the opposite wafer surface to the heat sink. This flow of heat through the wafer creates the temperature gradient that makes the TGZM process possible. A typical device useful for the commercial implementation of the TGZM method is U.S. Patent No. 4,221,956
It is stated in the specification of the issue. SUMMARY OF THE INVENTION In accordance with the present invention, recrystallized A method of forming one or more regions of single crystal semiconductor material is provided. This method is suitable for the manufacture of devices that require sharp boundaries (ie, interfaces) between the host material and the adjacent region of recrystallized semiconductor material. In particular, it is particularly useful when it is required to partially move the molten zone in the matrix, so as to form a sharp boundary at the interface between the matrix semiconductor material and the recrystallized semiconductor material in the direction of movement. The utility, stability and controllability of such methods are
As a driving force to move metal-containing melt,
It originates from the use of a potential gradient instead of a temperature gradient in the temperature gradient zone melting method (TGZM method). Since the metallurgical situation regarding the movement of metal-containing melts is essentially the same whether the driving force is a potential gradient or a temperature gradient, all the forms that can be formed in semiconductor crystals by the TGZM method are
It is anticipated that PN junctions may also be formed by the electromigration method of the present invention. Describing the steps of the method of the present invention, first of all,
Directions are determined for moving the metal-containing melt along respective paths through the matrix of single crystal semiconductor material located between the anode and cathode contacts. Second, one or more metallic solids are placed in contact with the matrix. The material constituting each metallic solid is (1) soluble in the base semiconductor material, and (2) forms a solution with the base semiconductor material that has a melting point lower than the melting point of the base semiconductor material. , and (3) When the work function of the base semiconductor material is φh, the algebraic sign of the quantity (φi − φh) and the algebraic sign of the contact area where the metal-containing melt should move and approach are the same. It is selected so as to satisfy the conditions of having the function φi.
The metallic solid material is any metallic element or alloy necessary to achieve the desired physical properties (e.g. resistivity, conductivity type) in the region of the recrystallized single crystal semiconductor material to be formed in the matrix. good. The solid metal material may also be a metallic element or alloy containing non-metallic elements such as boron or phosphorus. As used herein, the expression "placing a metal solid in contact with a matrix" means that the metal solid is placed completely inside the matrix (i.e., without contacting any surface). It covers both the case where the metal solid is initially placed on the surface of the matrix. Such metal solids may have a droplet-like, linear or planar network-like morphology. If the metal solid is linear, its long axis is substantially aligned with the crystallographically stable wire direction for the matrix and the metal to be transferred. Also, if the metal solid is reticulated, the long axes of the linear elements should be substantially aligned with the crystallographically stable wire directions for the matrix and the metal to be transferred. If a metallic solid is disposed on the exposed surface, such disposing is preferably done by means that produce a substantially oxygen-free metallic solid (eg, electron beam evaporation in vacuum). Third, anode and cathode contacts are added to the matrix. The aforementioned patent applications disclose, among other things, methods of forming contacts that are generally useful in the practice of electromigration methods and that differ from the simple pressure contact electrodes described herein. Fourth, the matrix and the metallic solids placed in contact with it are sufficient to produce a melt of metal-rich semiconductor material (i.e., a metal-containing melt) and substantially throughout the matrix. It is heated to a high temperature that is constant and uniform. Fifth, a potential gradient is applied across the matrix located between the anode and cathode contacts. Such potential gradients are generally set to be substantially parallel to the path for moving the one or more metal-containing melts. Finally, the driving force provided by the applied potential gradient causes the one or more metal-containing melts of the metal-rich semiconductor material to move simultaneously as molten zones along their respective paths in the solid matrix of semiconductor material. i.e. electromigration). The movement of each melt zone through the solid host semiconductor material results in dissolution of the liquid metal inclusion material in an amount up to the solid solubility limit of the metal-containing melt material in the host semiconductor material. A region of recrystallized single crystal semiconductor material is formed. Each region of the recrystallized single-crystal semiconductor material has a resistivity that differs from at least that of the parent semiconductor material, with the resistivity depending primarily on the material properties of the metal-containing melt and the temperature at which the electromigration method is carried out. It is determined. Furthermore, by appropriately selecting the material of the metal-containing melt, it is possible to make the region of the recrystallized single-crystal semiconductor material have a conductivity type different from that of the host semiconductor material. In carrying out the method of the present invention, for example, the first
One or more melt zones can be moved through the matrix from a major surface to a second major surface. Alternatively, after stopping the melting zone between the two major surfaces, the melting zone is moved up to the first major surface, preferably along the same path as initially, by applying a backward potential gradient; This also makes it possible to form sharp boundaries between the parent treated and untreated semiconductor material, especially along the direction of movement. Furthermore, when the electromigration method of the present invention is carried out when the matrix of single-crystal semiconductor material is a thin film, the electromigration method of the present invention is performed on a bulk matrix of semiconductor material.
It has also been found that regions of recrystallized single-crystalline semiconductor material form 100 to 1000 times faster. DETAILED DESCRIPTION The present invention will be more clearly understood from the following description in conjunction with the accompanying drawings. First, referring to FIG. 1, it has been found that in the single crystal semiconductor material 2 through which a direct current is passed, the metal-containing melt 1 moves toward the anode 3 or the cathode 4 along the flow line of the current. The movement of the metal-containing melt 1, as represented by the arrows in FIG. 1 and the following figures, reflects the flux of molten semiconductor atoms ( This is caused by flux). In order to create such a flux of semiconductor atoms, the atoms of the semiconductor material 2 must be
at the metal-containing melt 1 , while the dissolved semiconductor atoms are deposited at the rear surface 6 . For purposes of the present invention, since the direct current flow is from cathode 4 to anode 3, the direction (i.e. polarity) of the potential gradient across semiconductor material 2 is also from cathode 4 to anode 3. In this case, the metal-containing melt 1 is preferably arranged along the direction (ie, the path) of the potential gradient. Such metal-containing melt 1
In response to the potential gradient, the electromigration occurs substantially along the above-mentioned path towards the cathode 4 or the anode 3. In the region 7 of recrystallized monocrystalline semiconductor material deposited behind the electromigrating metal-containing melt 1,
Traces of the material (ie metal, non-metal or a combination thereof) comprising the metal-containing melt 1 remain. The remaining amount of such material is up to the solid solubility limit of the material of the metal-containing melt 1 in the semiconductor material 2. Since traces of the material of the metal-containing melt 1 are present in the region 7, the resistivity of the region 7 is equal to that of the semiconductor material 2.
The resistivity will be different from that of If the material constituting the metal-containing melt 1 is an electrically active dopant (also called an impurity) with respect to the semiconductor material 2, the region 7 can be of N-type or P-type depending on the valence state and concentration of the dopant. Become. If the P-type metal-containing melt 1 is electromigrated in the N-type semiconductor material 2, or if the N-type metal-containing melt is electromigrated in the P-type semiconductor material, A PN junction 8 can be formed. Since the electromigration rate of a metal-containing melt is orders of magnitude higher than the solid-state diffusion coefficient of the same metal, the electromigration of a metal-containing melt is a matter of minutes rather than days.
Provides a means for forming a bond. Even within the same semiconductor material, not all metal-containing melts will electromigrate in the same direction in response to a given potential gradient. As discovered in the course of completing the invention and schematically illustrated in FIG.
The metal-containing melt 9 of metal A, which has a work function larger than 4.55 electron volts (eV), moves electrically toward the anode 3 in silicon, whereas the metal B, which has a work function smaller than 4.55 eV, moves electrically toward the anode 3 in silicon. The contained melt 10 moves towards the cathode 4. Table 1 below shows the electromigration directions of nine metals in silicon and their work functions.
Antimony does not electromigrate in silicon because it has almost the same work function as silicon. However, antimony will also electromigrate in semiconductor materials that have a work function that is substantially different from that of silicon (eg, germanium or gallium arsenide).

【table】

[Table] Table 1 summarizes the data obtained during the process of completing the present invention. To obtain these data, an electromigration device 20, schematically shown in FIG. 3, and a standard diffusion furnace (not shown) were used. The device 20 is a 1/4 diameter cylindrical ceramic support 12 supported in two locations by a 2 inch (5.08 mm) diameter and 3 inch (7.62 cm) long cylindrical ceramic support 12.
It consisted of two molybdenum rods 11 facing each other with a diameter of 6.35 mm. To obtain structural rigidity, the molybdenum rods 11 in the ceramic support 12 are 1/8
The tube was inserted into a stainless steel tube 13 having a wall thickness of 2 inches (3.175 mm), an inner diameter of 2 inches (5.08 mm), and a length of 20 inches (50.8 cm). To achieve additional support, the molybdenum rod 11 was secured on the exterior of the stainless steel tube 13 with a clamp 14 made of an inert, non-conductive, rigid material. 1/8 inch (3.175 mm) in diameter as shown in Figure 3A below.
A replaceable molybdenum rod 15 (mm) was used as an electrode contact to the silicon specimen 16. During the high-temperature electromigration experiment conducted in the process of completing the present invention, contact pressure was applied between the molybdenum rod 15 and the silicon specimen 16 by a tungsten spring 17. Note that a ceramic piece 18 was placed between the springs 17 for electrical insulation. A test piece 16 measuring 4 cm x 1 cm x 1 cm was cut from an N-type silicon ingot having a resistivity of 10 Ω·cm. The long axis of such a rectangular test piece 16 is <111
> direction, and the top and side surfaces were perpendicular to the <110> and <112> directions, respectively. The upper surface of the test piece 16, that is, (11
0) From the surface, test piece 1 was prepared using ultrasonic drilling technology.
Two holes reaching the center of 6 were formed. Third
As shown in Figure A, these holes were symmetrically placed 1.5 cm from both ends of the specimen 16.
The larger hole 19 with a diameter of 2 mm is used as the thermocouple insertion hole, while the smaller hole 2 with a diameter of 0.75 mm
1 was used as a metal-containing melt insertion hole. The top and side surfaces of the specimen 16 were polished to a 1μ finish, and reference lines were drawn on the top and side surfaces using a diamond scriber so that the progress of electromigration could be observed by infrared transmission microscopy. Ta.
A metal wire with a diameter of 0.70 mm to be moved was inserted into the smaller hole 21. The specimen 16 was then placed in a 1100° C. annealing furnace (not shown) flushed with H ₂ gas for 3 hours to produce a metal-rich semiconductor solution for electromigration as a metal-containing melt. Thereafter, the specimen 16 was cooled and then removed from the annealing furnace. For testing electromigration parameters, the end of the device 20 holding the silicon specimen 16 and approximately half of the stainless steel tube 13 was inserted into a diffusion furnace (not shown) and heated to the electromigration temperature. . To prevent oxidation of the apparatus 20, 95% N ₂ +5% H ₂ was flushed through the furnace. After reaching temperature equilibrium, a constant DC current ranging from 3 to 30 amperes is applied to the
Flowed into silicon specimen 16 for ~240 hours. A fixed temperature of 400-1100°C was used and the temperature was continuously monitored by a chromel-alumel thermocouple inserted into the thermocouple insertion hole 19. After electromigration, device 20 was pulled out of the diffusion furnace and allowed to cool. By removing the test piece 16 and taking an infrared transmission micrograph, the moving distance of the metal-containing melt in the silicon test piece 16 was measured. The distance traveled was measured on a micrograph with reference to a reference line drawn on the surface of the silicon specimen 16 with a diamond scriber. The current is
Packard) 6269B constant current DC power supply (not shown). The reason you need a constant current power supply is
The resistivity of the silicon test piece 16 suddenly changed depending on the temperature. To maintain the 3 to 30 amperes of direct current used in the process of completing this invention, each approximately 1
A voltage drop of ~6 volts was required. The majority of these voltage drops occurred at the various electrical contacts rather than within the silicon specimen 16 itself. Joule heat, an external heat source (eg, a furnace), or a combination of both can be used to heat the specimen to a predetermined electromigration temperature. Although the use of Joule heat is the simplest and ultimately provides the highest rate of electrical transfer, it has the poorest controllability because the applied voltage provides both Joule heat and potential gradient. Controllability is best if external heating is performed while controlling the movement of the molten zone using a potential gradient, and it is also possible to create an in-and-out joint with a small depth from the surface where the movement starts. Ideal for forming. In general, whether the metal-containing melt moves toward the anode or the ^cathode , its electromigration velocity (V) is a linear function of current density unity (i.e., , it was found that if α is a constant of proportionality, then V=αI holds true. As a result, it was found that when electrotransfer and heat transfer methods were performed at the same nominal temperature, electrical transfer proceeded at the same speed as heat transfer at a relatively low current density of about 5 amperes/ ^cm2 . did. Figure 4 shows the direction of electromigration and the electromigration velocity/current (the quotient of the electromigration velocity by the current density) as a function of the electromigration temperature for eight types of metal-containing melts electromigrated in silicon. A plotted graph is shown. Theoretically speaking, metal-containing melts are capable of electromigration over a temperature range that limits the metal-semiconductor eutectic point and the semiconductor melting point. The electromigration rate/current for a metal-containing melt is determined by the Arrhenius temperature equation V/I=(V /I) ₀ exp(-Q/RT) Follows (1). In formula (1), V is the electric transfer speed (cm/sec),
I is the current density (ampere/cm ² ), Q is the activation energy (kcal/mol), R is the gas constant, T is the absolute temperature at which the electromigration method is carried out, and (V/I) ₀ is the pre-exponent. coefficient (cm ³ /clone). For the eight metals tested, activation energies ranged from a minimum of 14.5 kcal/mole for gallium to a maximum of 26.1 kcal/mole for palladium. The activation energies and pre-exponential coefficients for eight types of metals that can be electromigrated in silicon are shown in Table 1. Using the data in Table 1, an analytical study of the mechanisms involved in electromigration was performed. Details of such research can be found in The Journal of Applied Physics.
Applied Physics) Volume 51, No. 12 (December 1980)
It is reported in the article by Anthony on pages 6348-6355 and 6356-6365. To summarize these papers, the main driving force that brings about the electromigration of metal-containing melts is electron-
It turned out to be a momentum exchange between atoms, the so-called "electron wind effect." Furthermore, the direction of electromigration in silicon is correlated with the work function of the metal constituting the metal-containing melt, that is, a metal-containing melt consisting of a metal with a work function greater than 4.55 eV will move toward the anode. It was found that metal-containing melts made of metals with a work function smaller than 4.55 eV electromigrate toward the cathode, whereas they used to undergo electromigration. Finally, the relative order of the electromigration rates of the metal-containing melt in the anodic migration group and the cathodic migration group is determined by the work function such that no electromigration occurs (i.e., the work function of the host material that moves the metal-containing melt). This could be explained by the magnitude of the deviation of the work function φi of the metal-containing melt with respect to the value of φh) and factors related to the solubility and mass conservation of silicon in the metal-containing melt. Regarding a metal-containing melt moving through a solid host material, the flux J of host atoms passing through the metal-containing melt is expressed by the formula J=V(C _S −C _L ) (2) (where C _L is the concentration of host atoms per unit volume in the metal-containing melt, C _S is the concentration of host atoms per unit volume, and V is the migration velocity). This equation expresses the well-known law of conservation of mass. The above equation (2) is expressed by the well-known Einstein equation J=Mf( _CL ) (3) (where M is the mobility of the host atom in the metal-containing melt, and f is the mobility of the host atom in the metal-containing melt. in combination with the driving force acting on the host atom, and J and C _L are as defined above, we obtain the following equation: V=MfC _L /C _S −C _L (4) It was found that the velocity V of the electro-migrating metal-containing melt is proportional to the current density I and the deviation of the work function φi of the metal-containing melt with respect to the work function φh of the base material. Therefore, the formula V=βI△φ (5) (where △φ is equal to φi−φh and β is a constant) holds true. Comparing equations (4) and (5), it can be seen that the equation f=γI△φ (6) (where γ is a constant of proportionality) holds true. By combining equation (6) and equation (4), the following equation can be obtained. V/I=k(C _L /C _S −C _L )△φ (7) In the formula, k is a constant that can be derived from Figure 4 and Equation (7), and is 10 ⁺⁶ in the CGS unit system. Equal to cm ³ / (coulomb eV). Therefore, as can be seen from Table 1, if the value of V/I is negative, the metal-containing melt moves toward the cathode, and if the value of V/I is positive, the metal-containing melt moves toward the anode. There will be electric mobility in the future. In the manufacture of certain semiconductor devices, it may be desirable to use a metal-containing melt that electromigrates toward the anode and simultaneously dopes the silicon into a P-type. However, as can be seen from Table 1, the pure P-type dopants (Al, Ga, and In) all migrate toward the cathode. Therefore, in order to obtain a metal-containing melt that moves toward the anode and at the same time converts the silicon into P-type, it is necessary to prepare an alloy composition with an appropriate work function, such as a suitable proportion of Al.
and Pd. In order to obtain suitable metal-containing melts in these and other cases, information regarding the direction of electromigration of metal-containing melts consisting of two or more metals was sought. A priori, since an alloy consisting of two or more metals obeys the mixture rule, its overall work function φ ^* _i is given by the formula φ ^* _i =X _A φ _A +X _B φ _B +X _C φ _C +… ( 8) (where φ _A , φ _B , φ _C etc. are the work functions of pure elements, and X _A , X _B , X _C etc. are the atomic fractions of the elements constituting the alloy) It is expected that If the overall work function of such an alloy is greater than 4.55ev, the metal-containing melt is considered to electromigrate toward the anode in the silicon. On the other hand, the overall work function of such an alloy is 4.55eV
If it is smaller, the metal-containing melt is considered to be electromigrated in the silicon toward the cathode. In order to experimentally investigate the hypothesis that such a mixture rule holds, the direction of movement of a series of gold-silver alloys was measured. There are several reasons for choosing gold and silver. First, these noble metals are chemically similar and have nearly identical atomic radii. Moreover, because they produce solid and liquid solutions continuously, experiments are not complicated by extra factors such as eutectic mixtures, intermetallic compounds, or liquid immiscibility gaps. Of particular importance is the fact that the metal-containing melt consisting of these two noble metals moves in opposite directions. Furthermore, based on the present invention's findings regarding the electromigration of elements in silicon,
A temperature (884° C.) can be selected such that the absolute values of the moving speeds of the metal-containing melt consisting of gold and silver in opposite directions are equal. Finally, work functions for gold-silver alloys of various compositions are available in the literature, making it possible to compare the direction of electromigration of a particular alloy with its work function. By use of the apparatus 20 and method as described above, eleven gold-silver alloys were prepared with silver contents ranging from 0% to 100% in 10% steps (i.e., 0% Ag, 10% Ag, 20% Ag, etc.). , placed in separate metal-containing melt insertion holes in a single silicon specimen 16 and electrotransferred at 884°C.
FIG. 6 shows a graph in which the electromigration rate of these 11 types of gold-silver alloys is plotted against the composition. It has been found that alloys containing more than about 21% silver electromigrate towards the cathode, whereas alloys containing less than about 21% silver electromigrate towards the anode. To explain the direction of electromigration of gold-silver alloys in silicon, the Journal of
The Electrochemical Society
Journal of the Electrochemical Society) No.
As shown in Anthony's paper published in Vol. 128, No. 12 (March 1981), pp. 597-600, the work function of the metal constituting the metal-containing melt and the experimentally obtained null−point
A simple dilute solution theory constructed using the difference in work functions can be used. To summarize the above paper, Figure 7 shows gold-
This is a graph plotting the work function of a silver alloy against its composition. Based on the 4.55 eV zero point work function for silicon, the electromigration zero point composition of the gold-silver alloy is predicted to be 60% Au-40% Ag. However, since the electromigration velocities of metal-containing melts of pure gold and pure silver are equal and opposite at the above electromigration temperatures, simple linear extrapolation shows that the zero point composition is 50 %
It is shown that it becomes Au-50%Ag. However,
According to experiments carried out in the course of developing the present invention, a shift of the zero point towards gold-rich compositions was found. Qualitatively speaking, the shift of the zero point toward the gold-rich side means that the work function of the gold-silver alloy is not linear, and the change on the gold-rich side of the work function-composition diagram is more This was expected due to the fact that the rate of change is much steeper than that on the rich side. Thus, as far as the zero shift toward the gold-rich side is concerned, there is a qualitative agreement between the zero predicted from the work function of the gold-silver alloy and the actually observed zero. Admitted. However, quantitatively, 60% Au
The predicted zero point composition of −40%Ag is 79%Au
This does not match the measured zero point composition of -21%Ag. Such quantitative discrepancies are not surprising. This is because, of course, the work function of a pure alloy at room temperature is not equal to the work function of the same alloy liquid containing molten silicon at a temperature of 884°C. Considering the quantitative discrepancy between the direction of electromigration and the work function for gold-silver alloys, an almost perfect correlation was found between the direction of movement and the measured work function for pure metal-containing melts. It can be said that this is rather surprising. Based on the findings of the present invention, it is possible to predict the direction of electromigration of metal-containing melts. Alternatively, adjusting the composition of the metal-containing melt moves it in the desired direction, especially if the direction of the potential gradient is fixed by external factors (e.g. the properties of the host semiconductor material). It also becomes possible. Looking at FIG. 8, it can be seen that the type of impurity (or dopant) that imparts N-type to the matrix 30, for example.
A matrix 30 made of a single crystal semiconductor material containing atoms
A first example is shown. Assume that it is desired to form a P-type region 34 of recrystallized single crystal semiconductor material by electromigrating a metal-containing melt 32 toward the cathode 4, as shown. To do this, the algebraic sign (+ or -) of the difference obtained by subtracting the work function φh of the semiconductor material of the base body 30 from the work function φi of the material of the metal-containing melt 32 and the metal-containing melt 32 should move to approach each other. A material selected such that the algebraic sign (- in this case) of the contact portion (in this case, the cathode 4) is equal to that of the contact portion (in this case, the cathode 4) may be used as the metal-containing melt 32. In order to satisfy the conditions in the first case as described above for silicon, aluminum, indium or gallium may be used as the material of the metal-containing melt 32, as can be seen from Table 1. Tin and silver also electromigrate towards the cathode 4 and form recrystallized regions 34 having a resistivity different from that of silicon. However, these do not confer P- or N-type. If only the contact parts 3 and 4 are reversed (that is, if electric movement is performed toward the anode) while keeping the conditions shown in Fig. 8, the second
Examples are obtained. In this case, the metal-containing melt 32
The material must be chosen such that the algebraic sign of (φi − φh) is positive (+). In order to satisfy these conditions for silicon, palladium, gold, or copper can be used, as shown in Table 1, but all three metals are N
or P-type dopant. To meet these new requirements, alloys of palladium, gold or copper with P-type dopants (e.g. aluminum, indium or potassium) can be prepared, or palladium, gold or copper can be replaced with non-electromobile non-metallic P-type dopants. shaped dopants (e.g. boron)
must be used as a carrier for Note that examples of non-electromobile non-metallic N-type dopants include phosphorus and arsenic. In many cases, difficulties such as those seen in the second case described above in connection with FIG. 8 can be avoided by selecting the polarity of the potential gradient (i.e., reversing the anode and cathode contacts). Can be done. However, reversal of polarity is not possible in many cases, and thus polarity constitutes an additional constraint on electromigration methods. For example, in FIG.
Metal-containing melts 32 and 40 are shown forming P-type recrystallized regions 34 and N-type recrystallized regions 42, respectively, by electromigration simultaneously toward . In this case, metal-containing melts 32 and 4
All zero materials must be chosen such that the algebraic sign of (φi - φh) is negative. Although reversing the polarity may facilitate the selection of materials for the metal-containing melts 32 and 40, the polarity is a constraint. In the case of FIG. 10, a matrix 30 of single crystal semiconductor material
is a composite (i.e., P-type layer 44 on N-type layer 46)
). Since the interface 48 between layers 44 and 46 is a PN interface, the potential gradient can only be set as shown in FIG. Therefore, the metal-containing melt 32 is transferred to the P-type layer 4.
4 to form an N-type recrystallized region 34 or "sinker" by electromigration to the bonding surface 48 in
The material of 2 must be chosen such that the algebraic sign of (φi−φh) is negative. From the point of view of selecting the material of the metal-containing melt 32, in this case the eighth
The situation is similar to the second case described above in connection with the figure. Even if the conditions of FIG. 10 are changed so that layer 44 is N-type, layer 46 is P-type, and recrystallized region 34 is P-type, the potential gradient must be reversed;
Therefore, electromigration will occur toward the anode. Therefore, the difficult situation in the second case described above in connection with FIG. 8 still exists. However, according to the method of the invention it is also possible to adjust the composition of the material of the metal-containing melt such that electromigration can occur. Regardless of whether the driving force is a potential gradient or a temperature gradient, the metallurgical situation for the movement of metal-containing melts is essentially the same, so that all forms that can be formed in semiconductor crystals by the TGZM process are
Theoretically, a PN junction can also be formed by the electromigration method of the present invention. To gain a thorough understanding of the TGZM technology applicable to the practice of the present invention (e.g., preparation of semiconductors and metal-containing melt materials for transfer of metal-containing melts) and the products obtained thereby. In addition to the aforementioned U.S. patent specifications, U.S. Pat.
No. 3902925, ``Fast Heat Transfer Method for Manufacturing Deep Diodes,'' No. 3898106, entitled ``Stabilized Droplet Method for Manufacturing Deep Diodes with Uniform Electrical Properties.'' No. 3,899,361, ``Heat Transfer of Metal-Rich Liquid Wires in Semiconductors,'' 3,899,962, and 4,006,040, ``Manufacturing of Semiconductor Devices.'' However, those skilled in the semiconductor art will appreciate that the present invention provides benefits that are not necessarily achieved in practicing the TGZM process as exemplified by the above-mentioned US patents. For example, by utilizing the unique discovery of the present invention, it is possible to reach deep inside from the surface of the matrix on the semiconductor side, and to create an extremely clear gap between the matrix semiconductor material and the recrystallized region when viewed along the electromigration direction. It is possible to form a diode with a junction surface of Such a sharp bonding surface is obtained because the direction of the slope of the applied potential can be rapidly reversed. Furthermore, because the rate of electromigration is generally greater than the rate of heat transfer at a given temperature, electromigration methods can be performed more rapidly. the result,
Since the time for the dopant atoms to diffuse into the host semiconductor material is reduced, a sharp bonding surface is formed even in the direction perpendicular to the direction of movement. In the process of perfecting the present invention, three phenomena were discovered that may affect the driving force due to the potential gradient of the electromigration method. These are the formation of rectifying junctions, current concentration in the metal-containing melt, and current concentration in the doped recrystallized region. The formation of a rectifying junction depends on the temperature and conductivity type (i.e., P type, N
form or essence). At sufficiently high temperatures, both the host semiconductor material and the recrystallized region formed behind the electromigrating metal-containing melt exhibit an intrinsic conductivity type. As the temperature decreases and reaches a first transition temperature, the recrystallized region begins to exhibit an impurity conductivity type (ie, P-type or N-type), while the host semiconductor material continues to exhibit an intrinsic conductivity type. The reason why the recrystallized region exhibits the impurity conductivity type before the host semiconductor material is that either only the former contains dopant atoms, or that the concentration of dopant atoms in the former is higher than in the latter. A typical case is schematically illustrated in FIG. In this case, the metal-containing melt 60 electromigrates toward the cathode 4 to form a P-type recrystallized region 66, but the base semiconductor material 62
remains an intrinsic semiconductor. As will be appreciated by those skilled in the semiconductor art, the interface 64 between the recrystallized region 66 and the host semiconductor material 62 is reverse biased. Once such a rectifying junction is formed, the applied current 68 will bypass the metal-containing melt 1, as schematically shown in FIG. As a result, a loss of driving force arises due to the potential gradient, and eventually electrical migration stops. As the temperature decreases further, the host semiconductor material also changes from an intrinsic conductivity type to an impurity conductivity type at a second transition temperature. A typical case is schematically shown in FIG. In this case, the rectifying junction is caused by the metal-containing melt 70 being electromigrated through the P-type matrix 72 to form the N-type recrystallized region 76 .
It has the form of an inter-semiconductor junction (ie, a Schottky junction). For reverse biased interfaces involved in the electromigration of P-type and N-type metal droplets in intrinsic, P-type and N-type semiconductor materials, as disclosed in the aforementioned U.S. Patent Application No. 272,801. Approximately 16 different cases have been identified. In six of these cases, forming the anode and cathode contacts as described in the above-mentioned U.S. patent application solves the rectifying junction problem and improves electromigration doping capability and electromigration temperature. The range can be expanded. It is believed that if there is a large difference between the conductivity of the host semiconductor material and the conductivity of the metal-containing melt, the current density in the metal-containing melt can be different from the current density in the host semiconductor material. . Moreover, such a current density mismatch could result in an increase in migration speed due to current crowding effects, thereby interfering with precise control of the electromigration process. However, The Journal of Applied
Physics (The Journal of Applied
Physics) Volume 51, No. 12 (December 1980), 6348~
As shown in the aforementioned paper on page 6355, the actual current density in the metal-containing melt is equal to the current density applied to the host semiconductor material, so no significant speed-enhancing effect occurs. . Not only the difference between the conductivity of the host semiconductor material and the metal-containing melt, but also the difference between the conductivity of the host semiconductor material and the metal-doped recrystallized region that forms behind the electromigrating metal-containing melt. There is also a size difference between the conductivity and the conductivity at most temperatures. This conductivity difference between the host semiconductor material and the recrystallized region concentrates the applied current into the metal-containing melt. Such an effect is illustrated in FIG. 14 for a gallium-rich metal-containing melt electromigrated in silicon. As can be seen from FIG. 14, such an effect becomes more and more pronounced as the electromigration temperature decreases.
This effect is enhanced by the length of the doped recrystallized region, so that for metal-containing melts that move deep into the host semiconductor material, as is the case with finger diodes, the current into the recrystallized region is Concentration is one factor. However, as was discovered in the course of developing the present invention, the mismatch between the conductivity of the host semiconductor material and the recrystallized region can be compensated for experimentally. Such compensation is possible because it was simultaneously discovered that the penetration distance of liquid metal inclusions moving from the surface into the interior of the host semiconductor material is directly proportional to the coulomb/cm ² value in the host semiconductor material. by. That is, a calibration curve can be constructed that plots the entry distance against the coulomb/cm ² value for a range of electrotransfer temperatures. Such a calibration curve would already take into account the effects of current concentration in the doped recrystallized regions. In fact, once a calibration curve has been developed, it is possible to produce finger diodes with very precisely controlled penetration distances in the host semiconductor material. Since the amount of current used can be measured to about ±10 coulombs, the droplet penetration distance can be controlled to about 1/100 micron. To give a practical example, if we know from the calibration curve that 25×10 ⁴ clones/cm ² can electromigrate a gallium-rich droplet 100 microns into silicon at 900°C, then Finger diodes can be manufactured. First, the surface on the starting side is at a positive potential with respect to the surface on the opposite side, and ^the
A current of clones/cm ² is passed into the mother's body. The gallium is then charged by reversing the direction of the current (i.e. with the starting surface at a negative potential with respect to the opposite surface) and applying a current of slightly more than 25 x ¹⁰ clones/ ^cm2 . It is sufficient to return the enriched droplets to the starting surface. As an example of the application of the novel discovery that metal-containing melts can be electromigrated in a bulk host semiconductor material, it has been found that electromigration methods can also be carried out in thin films. Turning to FIG. 15, a composite consisting of a substrate 80 and a thin film 82 of substantially uniform thickness is schematically illustrated. Thin film 82 is comprised of a material deposited along at least a portion of substrate 80 by vapor deposition or other methods. Further, the substrate having the upper and lower surfaces located back to back and the peripheral end surface connecting them can be made of a metal (e.g., copper) or a semiconductor (e.g., silicon, germanium, a Group-3 compound, or a Group-3 compound).
In order to concentrate the current in the thin film 82, the substrate 80 must not be electrically conductive. Or again,
Substrate 80 is used to limit the current to only in thin film 82.
At the interface 84 between the top surface and the bottom surface of the thin film 82
A thin insulating film of material such as SiO ₂ , Al ₂ O ₃ or diamond may be disposed. Suitable materials for thin film 82 include silicon, germanium, gallium phosphide, gallium arsenide, indium antimonide, cadmium telluride, and zinc sulfide. Thin film 82 is silicon and substrate 80
is saphire, then thin film electromigration techniques can be performed on silicon-on-sapphire materials known to those skilled in the semiconductor arts. The cathode 4 and the anode 3 are respectively connected to the thin film 82 via contact portions 85 and 86 which are arranged substantially facing each other on the outer peripheral end surface of the thin film 82 . A suitable method for forming contact portions 85 and 86 is disclosed in the aforementioned US patent application Ser. No. 272,801. Next, one or more metal solids 88 comprising the metal to be electromigrated are applied to the thin film 82.
is placed on the exposed surface 90 of. Alternatively, the metal to be electromigrated may be placed during the operation to apply contacts 85 and 86, as described in the aforementioned US patent application Ser. No. 2,72,801. Thereafter, the composite consisting of membrane 82 and substrate 80 is heated to the electrotransfer temperature by any of the methods described above. For thin films 82 having a thickness of about 1 mil (0.0254 mm) or less, metallic solids 8
8 rapidly diffuses throughout the thickness of the thin film 82. Once the metal solid 88 is liquefied and distributed throughout the thickness of the thin film membrane 82, electromigration is initiated by applying a DC potential between contacts 85 and 86. For films 82 thicker than about 1 mil (0.0254 mm), if it is desired to achieve penetration through the thickness of the film 82 in a reasonable amount of time,
The metal solid 88 must be forced through the entire thickness. At least in such cases, the opportunity is available to carry out the temperature gradient zone zone melting method and the electromigration method simultaneously. To this end, a composite consisting of a membrane 82 and a substrate 80 and having a metal solid 88 to be electromigrated is placed in a suitable device for establishing and maintaining a finite temperature gradient through the membrane 82 and substrate 80.
In that case, the exposed surface (i.e., the upper surface) of the thin film 82
90 is maintained at a lower temperature than the exposed surface (ie, bottom surface) 92 of the substrate 80. The temperature in the temperature gradient at least in the portion passing through the thin film 82 is lower than that of the metal solid 88.
If set and maintained above the melting point of , the metal solid 88 will melt to form a metal-containing melt 88 . Due to the presence of the temperature gradient, the metal-containing melt 88 moves along the thickness of the thin film 82 until it reaches the interface 84 between the thin film 82 and the substrate 80 . Then, when an electromigration operation is performed, the metal-containing melt 88 moves to the contact portion 8.
5 and 86 across the thin film 82. At the same time, by the TGZM method,
The metal-containing melt 88 is maintained throughout the thickness of the thin film 82. In this case, the temperature gradient is substantially perpendicular to the potential gradient. It has been found that the electromigration velocity of a liquid zone in a thin film is significantly increased over the maximum velocity in a bulk solid, as shown by FIG. To obtain Figure 16, the velocity of an aluminum metal-containing melt was measured by electrotransferring it at various temperatures in a thin (4μ) silicon layer of a commercial silicon-on-sapphire (SOS) wafer. Ta. Also plotted in FIG. 16 is the data obtained from FIG. 4 regarding the electromigration of aluminum in bulk silicon. As can be seen from FIG. 16, the electromigration rate in a thin film is about two to three orders of magnitude higher than that in a bulk solid. For purposes of illustration and not limitation, the rate of electromigration in thin films is similar to that in bulk solids.
It is 100 to 1000 times. This is primarily due to the fact that the intersection lines of the liquid zone with the free surface provide abundant nucleation sites (e.g. atomic steps and ledges), whereas the liquid zone inside the crystal does not. This comes from the fact that That is, when a liquid band intersects a free surface, as in thin film electromigration, there is no nucleation barrier to movement of the liquid band. Second, current densities in thin films can be orders of magnitude greater than current densities in bulk materials without causing melting. That is, extremely large current densities (i.e., current densities on the order of 10 ⁶ amperes/cm ² ) are applied to the thin film, since the Joule heat that would cause melting of bulk materials is removed through the underlying substrate. It is possible. In this way, the speed of electrical transfer is essentially (approximately 10 ² ~
Combined with the fact that the current density can be increased (by about 10 ³ to 10 ⁴ times), the rate of electrotransfer in thin films is 10 ⁵ to 10 ⁷ times larger than ⁱⁿ bulk materials. It is possible to take a value as large as that (ie, a value of about 300 to 3000 cm/hour). Processing costs for thin film electromigration are expected to be less than processing bulk materials by TGZM or electromigration for two reasons. 1st
The reason for this is that although an extremely high current density is used, the actual amount of current required is not large because the thin film along the current path is extremely thin. A second reason for the lower cost of thin film electromigration is that it does not require as much equipment capacity as electromigration in bulk materials.
The cost of processing equipment can be reduced. Although the present invention has been described above with particular reference to silicon semiconductor materials, the present invention is a novel technique for moving a metal-containing melt in a host semiconductor material using a controllable and reversible potential gradient as a driving force. may also be used with other materials known in the art such as germanium (Ge), gallium phosphide (GaP), gallium arsenide (GaAs), indium antimonide (InSb), cadmium telluride (CdTe) and zinc sulfide (ZnS). Can be widely applied. Furthermore, while the invention has been described in detail with respect to several preferred embodiments, changes may be made in form and detail without departing from the spirit and scope of the invention as defined by the following claims. It will be obvious to those skilled in the art that other modifications may be made.

[Brief explanation of drawings]

FIG. 1 is a schematic cross-sectional view of a metal-containing melt electrotranslocating within a matrix of semiconductor material in response to an applied potential gradient, and FIG. 2 is a schematic cross-sectional view of the same semiconductor material in response to the same applied potential gradient. FIG. 3 is a schematic cross-sectional view of an apparatus useful for carrying out the electromigration method; FIG. Fig. 3 is an enlarged schematic cross-sectional view of the circled area, and Fig. 4 shows the electromigration direction and electromigration speed/current relative to the electromigration temperature for eight types of metal-containing melts that electromigrate in silicon. Fig. 5 is a semi-logarithmic graph plotting the electromigration rate/current versus electromigration temperature for a silver-rich metal-containing melt electromigrating in silicon; For a moving series of melts containing gold-silver alloys, the electromigration velocity/
Figure 7 is a graph in which the current is plotted against the alloy composition, Figure 7 is a graph in which the work function is plotted against the alloy composition for a gold-silver alloy, and Figure 8 is a graph in which the work function is plotted against the alloy composition for a gold-silver alloy. FIG. 9 is a schematic cross-sectional view of a metal-containing melt forming a region of P-type semiconductor material which has been recrystallized by electromigration through a matrix of N-type semiconductor material; FIG. FIG. 10 is a schematic cross-sectional view of two types of metal-containing melts forming regions of recrystallized N-type semiconductor material, respectively. FIG. 11 is a schematic cross-sectional view of a metal-containing melt forming an N-type region in a shaped layer. FIG. 12 is a schematic cross-sectional view of a metal-containing melt forming a region of shaped semiconductor material; FIG. 12 is a schematic cross-sectional view showing the current dispersion phenomenon caused by the formation of a rectifying junction; FIG. 13 is a schematic cross-sectional view of a Schottky-shaped rectifying junction; Figure 14 is a semi-logarithmic graph plotting the conductivity of intrinsic silicon and the conductivity of a gallium-containing silicon region formed by electromigration of a gallium-rich metal-containing melt against temperature. , FIG. 15 is a schematic diagram illustrating thin film electromigration technology, and FIG. 16 is a semi-logarithmic graph of electromigration velocity/current plotted against the reciprocal of temperature for aluminum electromigrated in thin silicon films and bulk silicon. . In the figure, 1 is a metal-containing melt, 2 is a matrix of single crystal semiconductor material, 3 is an anode, 4 is a cathode, 7 is a region of recrystallized single crystal semiconductor material, 8 is a PN junction, and 30 is single crystal semiconductor material 32 and 40 are metal-containing melts, 34 and 42 are regions of recrystallized single-crystalline semiconductor material, 60 is metal-containing melt, 62 is a base of single-crystalline semiconductor material, 64 is a bonding surface, and 66 is a recrystallized single-crystalline semiconductor material. a region of crystallized single crystal semiconductor material, 68 a current;
70 is a metal-containing melt, 72 is a matrix of single crystal semiconductor material, 76 is a region of recrystallized single crystal semiconductor material, 80 is a base body, 82 is a thin film, 84 is an interface, 85
and 86 represent a contact portion, and 88 represents a metal solid or a metal-containing melt.

Claims (1)

Claims: 1. Moving one or more metal-containing melts through the use of a potential gradient in a matrix of single-crystalline semiconductor material, thereby creating a resistivity that is at least different from that of the matrix semiconductor material. A method of forming one or more regions of recrystallized single-crystalline semiconductor material in said matrix having: (a) depositing said one or more metal-containing melts in said matrix located between an anode and an anode contact. determining a direction for movement along one or more paths; (b) (i) soluble in the host semiconductor material; When a solution having a melting point lower than that of the semiconductor material is generated, and (iii) the work function of the base semiconductor material is φh, the amount (φi − φh)
made of a material selected to satisfy the conditions that it has a work function φi such that the algebraic sign of is equal to the algebraic sign of the contact portion to which the one or more metal-containing melts should move and approach. placing one or more metal solids each in contact with the matrix;
(c) adding the anode and cathode contact portions to the matrix; (d)
subjecting the one or more metal solids to an elevated temperature that is sufficient to melt the one or more metal solids to form one or more metal-containing melts while being substantially constant and uniform throughout the matrix; heating a matrix; (e) establishing the potential gradient between the anode and cathode contacts;
and (f) moving the one or more metal-containing melts along the path in the matrix through the use of the potential gradient to cause the one or more regions of recrystallized single crystal semiconductor material to A method characterized by comprising steps of forming it in a state integrated with the matrix. 2. The method of claim 1, wherein at least one of the regions of recrystallized single crystal semiconductor material also has a conductivity type different from that of the parent semiconductor material. 3. The method of claim 1, wherein the metallic solid is placed inside the matrix. 4. The method of claim 1, wherein said heating step is accomplished by the use of an external heat source. 5. The method of claim 1, wherein the heating step is accomplished by utilizing Joule heat. 6. The method of claim 1, wherein the heating step is accomplished by a combination of an external heat source and Joule heat. 7. The one or more solid metal materials are gallium,
silver, copper, aluminum, gold, tin, indium,
The method of claim 1, wherein the material is selected from the group consisting of palladium and alloys thereof. Claim 7, wherein a non-electromobile dopant source selected from the group consisting of boron, phosphorus and arsenic is included in the one or more metal solids.
The method described in section. 9. The semiconductor material of claim 1, wherein the base semiconductor material is a material selected from the group consisting of silicon, germanium, gallium arsenide, gallium phosphide, indium antimonide, cadmium telluride, and zinc sulfide. Method. 10. The method of claim 1, wherein the matrix is a matrix thin film of single crystal semiconductor material disposed along at least a portion of the surface of the substrate. 11. Claim 10, wherein at least one of the recrystallized single crystal semiconductor material regions also has a conductivity type different from the conductivity type of the parent semiconductor material.
The method described in section. 12. The method of claim 10, wherein said heating step is accomplished by use of an external heat source. 13. The method of claim 10, wherein the heating step is accomplished by utilizing Joule heat. 14 Claim 10, wherein the heating step is achieved by a combination of an external heat source and Joule heat.
The method described in section. 15. Claim 10, wherein the material of the substrate is selected from the group consisting of copper, saphire, germanium, gallium phosphide, gallium arsenide, indium antimonide, cadmium telluride, and zinc sulfide. the method of. 16. The semiconductor material of the base thin film is a material selected from the group consisting of silicon, germanium, gallium phosphide, gallium arsenide, indium antimonide, cadmium telluride, and zinc sulfide. the method of. 17. The method of claim 10, wherein a thin insulating film comprised of one of silicon dioxide, alumina or diamond is disposed at the interface between the substrate and the host film. 18. Claim 10, wherein the one or more solid metal materials are selected from the group consisting of gallium, silver, copper, aluminum, gold, tin, indium, palladium and alloys thereof.
The method described in section. 19. The method of claim 18, wherein a non-electromobile dopant source selected from the group consisting of boron, phosphorous and arsenic is contained in the one or more metal solids. 20 moving one or more metal-containing melts through the use of a potential gradient in a matrix thin film of single crystal semiconductor material disposed along at least a portion of the surface of a substrate, thereby displacing at least the semiconductor material of said matrix thin film; A method of forming in said host thin film one or more regions of recrystallized single crystal semiconductor material having a resistivity different from said one.
determining a direction for moving one or more metal-containing melts along one or more paths in the matrix film located between the anode and cathode contacts;
(b)(i) is soluble in the semiconductor material of the host thin film;
(ii) When a solution having a melting point lower than the melting point of the semiconductor material of the base thin film is generated between the semiconductor material of the base thin film, and (iii) the work function of the semiconductor material of the base thin film is φh. , satisfies the conditions that the work function φi is such that the algebraic sign of the quantity (φi−φh) and the algebraic sign of the contact portion to which the one or more metal-containing melts should move and approach are equal. (c) affixing the anode and cathode contacts to the matrix membrane; (c) applying the anode and cathode contacts to the matrix membrane;
(d) The temperature of the upper surface of the base thin film is lower than the temperature of the lower surface of the substrate, and the temperature in the temperature gradient of the portion passing through the base thin film is higher than the melting point of the one or more metal solids. establishing and maintaining a finite temperature gradient through the host film and the substrate to form one or more metal-containing melts extending throughout the thickness of the host film;
(e) setting the potential gradient between the anode and cathode contacts so as to be substantially perpendicular to the thickness direction of the base thin film and substantially perpendicular to the temperature gradient; and (f) displacing one or more regions of recrystallized single crystal semiconductor material by moving said one or more metal-containing melts along said path in said host thin film through the use of said potential gradient. A method characterized by comprising steps of forming it in a state integrated with the base thin film. 21. The method of claim 20, wherein at least one of the regions of recrystallized single crystal semiconductor material also has a conductivity type different from the conductivity type of the semiconductor material of the host thin film. 22. Claim 20, wherein the material of the substrate is selected from the group consisting of copper, sapphire, germanium, gallium phosphide, gallium arsenide, indium antimonide, cadmium telluride, and zinc sulfide. the method of. 23. The semiconductor material of the base thin film is a material selected from the group consisting of silicon, germanium, gallium phosphide, gallium arsenide, indium antimonide, cadmium telluride, and zinc sulfide. the method of. 24. The method of claim 20, wherein a thin insulating film comprised of one of silicon dioxide, alumina or diamond is disposed at the interface between the substrate and the host film. 25. Claim 20, wherein the one or more metallic solid materials are selected from the group consisting of gallium, silver, copper, aluminum, gold, tin, indium, palladium, and the like. the method of. 26. The method of claim 20, wherein a non-electromobile dopant source selected from the group consisting of boron, phosphorous and arsenic is included in the one or more metal solids.