RU2208068C1 - Method of preparing silicon crystals with cyclic twin structure - Google Patents

Abstract

FIELD: crystal growing. SUBSTANCE: silicon crystals with cyclic twin structure appropriate for fabrication of semiconductor ingots or plates are grown by Chokhralsky method from melt onto seed with cyclic twin structure varying from basic three-seed structure formed by two coherent first-order twinning surfaces and one second-order twinning border to full cyclic twin structure formed by twenty coherent first-order twinning surfaces, four second-order twinning borders, six third-order twinning borders, and even number of additional first-order twinning surfaces in parallel to above-mentioned twenty surfaces. Growing is performed by adding to silicon melt additives selected from germanium, tin, and lead at concentration 1,0•10^-7 to 15% of the weight of silicon. EFFECT: increased length of grown crystals with cyclic twin structure and thereby increased productivity of growing these crystals. 1 tbl

Description

 The present invention relates to the field of production of semiconductor ingots and wafers that can be used, for example, in the manufacture of solar cells. In particular, the present invention relates to cyclically twin crystals of semiconductor materials, especially silicon, crystallizing in a diamond cubic lattice, and a method for their preparation.

 The main materials used for the manufacture of solar cells are dislocation-free silicon single crystals obtained by the Czochralski method, and multicrystalline silicon obtained by casting. The raw materials used in these methods are polycrystalline raw silicon semiconductor purity, as well as the waste of single-crystal silicon ingots intended for micro- and power electronics and obtained by the Czochralski method and crucible-free zone melting from polycrystalline raw silicon.

 A rough estimate shows that in order to satisfy the global demand for solar cells of ~ 1400 MW / year, approximately 84,000 tons of raw polycrystalline silicon raw material are needed to produce dislocation-free silicon single crystals by the Czochralski method and multicrystalline silicon ingots by casting. Polycrystalline raw silicon is, first of all, a raw material for producing silicon single crystals intended for microelectronics. The cost of this raw material is too high (~ 50 USD / kg) to use it to obtain the substrate material of solar cells. At the same time, it is impossible to use low-quality silicon, such as refined metallurgical silicon, to obtain dislocation-free silicon, due to the extreme sensitivity of the process of dislocation-free crystal growth to impurities (contaminants) and foreign particles, even if they are of the smallest size.

 To use cheap low-quality feedstock in the production of a substrate material for silicon solar cells with a sufficiently high efficiency It is necessary to develop a special method for producing silicon crystals with an acceptable structure and physical properties.

 Another objective in the production of solar cells is a significant reduction in the thickness of the used substrates (wafers) in order to reduce the cost of solar cells and use silicon crystals with a low lifetime of minority charge carriers for the manufacture of solar cells with high efficiency Despite the fact that silicon as a material is sufficiently solid, silicon single crystals are very fragile due to the presence of four {111} planes in them, which are cleavage planes that completely intersect the single crystal. This is the main reason why thin silicon wafers break very easily. Therefore, cutting silicon single crystals into very thin wafers with a high yield percentage is practically impossible. In this regard, it seems that coarse-grained silicon crystals with a regular structure could be a suitable material for solving this problem.

 The first attempts to develop a method for producing coarse-grained silicon crystals with a regular twin structure for use in solar cells were made by J. Martinelli and R. Kibizov in 1992 (see G. Martinelli, R. Kibizov "Growth of stable dislocation-free 3-grain silicon ingots for thinner slicing. "Appl. Phys. Letters, Vol. 62, June 21, 1993, pp. 3262-3263). The resulting material was a semiconductor crystalline silicon with three adjacent, sectorially located single crystal zones - the so-called three-grain silicon. In this work, the possibility of ultrafine cutting of three-grain silicon ingots into wafers is demonstrated and the possibility of their use for manufacturing highly efficient solar cells is shown.

 One of the methods for producing triple-grain silicon crystals is described in the patent of Germany 4343296 C2 (publ. 12.09.96, H 01 L 31/036). The process involves the preparation of seed crystals by sawing three regular octahedra from a silicon single crystal with all surfaces aligned along the {111} crystallographic planes; further, the cultivation of a two-grain ingot from a melt by using as a seed two prepared octahedra located relative to each other in a twin position and connected to each other by molybdenum wire; then sawing along the {111} planes of the prismatic sector from the grown ingot; inserting the third octahedron in the double position into the excised prismatic sector of the grown crystal and bonding them together with molybdenum wire; shortening the two-grain crystal to the length of the third inserted octahedron and, finally, growing a three-grain crystal from a silicon melt by means of a seed crystal thus prepared.

Through this patented process, three-grain crystals can be obtained, however, this process has several disadvantages:
1) It is very difficult to fabricate octahedral crystals and cut out the prismatic sector with surfaces having an exact crystallographic orientation of {111}, and it is also very difficult to mechanically connect the octahedral crystals and the octahedral crystal with the cut plane of the two-grain crystal in the exact twin position and with the exact coincidence of the crystal lattices. This circumstance implies the creation of stresses and structural defects in the grown ingot along the boundaries of twins.

 2) The use of octahedra as crystals in seed form is technically extremely difficult due to the large diameter to length ratio of approximately 2. In conventional silicon ingot technology, the seed crystals used have a diameter of approximately 12 mm and a length of 100 150 mm. It is also possible to use seeds with a length of 30-50 mm, but not shorter. But in the case of using bound octahedra as seeds with their length equal to 30-50 mm, their diameter will be 60-100 mm, which makes the growth process very difficult.

 3) The cultivation of three-grain ingots at somewhat higher drawing speeds than single-crystal ingots is possible due to the formation of so-called incoming angles formed by {111} planes at the exit points of twin boundaries at the crystallization front. As shown by R.S. Wagner (Acta Metallurgica. Vol. 8, 1960, pp. 57-60) and D.R. Hamilton and R.G. Seidensticker (Journal of Applied Physics. Vol. 31, 1960, pp. 1165-1168), these incoming angles form the regions (places) of the easiest nucleation. However, it was also shown in the above papers that the condition for self-reproduction of the incoming angles and rapid crystal growth is the presence of at least two or more closely spaced twin planes. Otherwise, the incoming corners wedge out and rapid growth stops.

The closest is the patent of Finland 106729 (publ. 30.03.2001, C 30 B 15/00), which proposed simple methods of manufacturing seed crystals, ingots and wafers with a cyclic twin structure and crystals with a cyclic twin structure from the base - three-grain, to the full - twenty-grain. Such twin crystals contain first-order coherent twinning planes, second-order twin boundaries and may contain third-order twin boundaries. The method of growing silicon crystals with a cyclic twin structure has a number of advantages, which include:
1) The possibility of radical improvements in growing technology (for example, the use of multiple semi-continuous growing), which reduces the cost of these crystals.

 2) The ability to use raw materials of low quality (low cost) without degrading the parameters of devices (solar cells) made on the basis of crystals with a cyclic twin structure, and reducing due to this cost.

 3) The possibility of ultra-fine cutting onto wafers of crystals with a cyclic twin structure due to the hardening of crystals by twin planes and boundaries and a reduction in cost due to this.

 However, the method for producing crystals with a cyclic twin structure, described in Finnish patent 106729, does not allow them to grow a sufficient length, which limits productivity and inhibits further cost reduction.

 It is known that first-order coherent twinning planes are not electrically active, since such twinning does not lead to the formation of dangling and distorted bonds. The twin boundaries of the second and higher orders are defective in nature. They arise, as a rule, as a result of the fusion of twin individuals. Such intergrowth leads to the formation of distorted and dangling bonds. Moreover, the higher the order of the twin boundary, the more defective this boundary is. In addition, the structure of twin boundaries of the second and third orders, in addition to crystallographic factors, depends on the growing conditions and may differ from the equilibrium structure. In this regard, despite the rather stable and reproducible growth of crystals with a cyclic twin structure, it is difficult to grow these crystals of large length without losing the regular structure. As a rule, as these crystals grow, dislocations from incoherent twin boundaries (second and third orders) are generated. As a result, starting from a length of ~ 250–300 mm, polycrystalline inclusions are formed in them. This fact limits the productivity of the process of obtaining crystals with a regular cyclic twin structure without polycrystalline inclusions.

 The aim of the present invention is to increase the length of the grown silicon crystals with a cyclic twin structure without polycrystalline inclusions and, therefore, increase the productivity of the process for producing these crystals.

 Another objective of the present invention is to develop a method for producing silicon crystals with improved properties for the manufacture of highly efficient solar cells based on them.

The specified technical result is achieved by the fact that in the method for producing silicon crystals with a cyclic twin structure by growing from a melt by the Czochralski method for seed with a cyclic twin structure from a basic - three-grain, formed by two coherent twinning planes of the first order and one twin boundary of the second order, to the full cyclic twin structure formed by twenty coherent twin planes of the first order, with four twin boundaries second order, with six boundaries of twinning of the third order and an even number of additional twinning planes of the first order parallel to the specified twenty, according to the invention, the growth is carried out with the introduction of additives selected from a number of germanium, tin, lead into the silicon melt, while the concentration of additives in relation to silicon is 1.0 • 10 ^-7 -15 wt.%.

 These and other objectives, together with the advantages of the present invention compared to known processes and materials, should become apparent from the following description and claims.

 The present invention is based on the idea of growing long (> 300 mm) silicon crystals from a melt in the <110> direction with a diverse cyclic twin structure: from a basic cyclic twin structure formed by two coherent twin planes of the first order and one twin boundary of the second order, to the full cyclic twin structure formed by twenty coherent twin planes of the first order, four boundaries of twinning of the second order and six boundaries twinning of the third order depending on the degree of supercooling of the melt. Moreover, to passivate incoherent twin boundaries, i.e., to reduce the likelihood of generation of dislocations by these boundaries and the formation of polycrystalline inclusions, passivating impurities with larger atomic radii than silicon, which are isovalent impurities, are introduced into the silicon melt (in the initial silicon charge) like germanium, tin and lead.

 The structure of twin boundaries in crystals with a diamond structure was studied by J. Kohn and J. Hornstroy (Kohn JA, Amer. Mineralogist, 41, 9/10, 778-784 (1956); Kohn JA, Amer. Mineralogist, 43, 3/4, 263-284 (1958); Hornstra J., Physica, 25 (6), 409-422 (1959); Hornstra J., Physica, 26 (3), 198-208 (I960)).

 It was shown in these works that crystals in the region of twin boundaries have distortions, increased interatomic distances, and dangling bonds, and that twin boundaries “look for” positions with minimal energy and can take a zigzag shape.

 In real growing processes, crystallization conditions differ from equilibrium conditions; therefore, twin boundaries cannot always find the most favorable positions in the energy sense, which should lead to an increase in the energy of the fields of mechanical stress formed by them. The relaxation of these stresses during growth is precisely the reason for the generation of defects in the crystal structure and the formation of polycrystalline inclusions.

 The introduction of additives with substances of atoms of other sizes into the melt allows the growing crystal to expand the possibilities of choice for the formation of twin boundaries with minimal energy and to enable the “healing” of already formed boundaries. Confirmation of the desire of the crystal to use impurity atoms to form boundaries with minimum energy is the fact of segregation of impurities in the region of grain boundaries.

 For the case of growing silicon crystals, isovalent impurities were chosen as such additives: germanium, tin, and lead, since their introduction into the crystal does not lead to an uncontrolled change in the concentration of charge carriers. These additives can be introduced into the crystal (melt) both separately and together.

The concentration values of these additives necessary for the formation of twin boundaries that are stable during growth depend on the specific growing conditions: growth process parameters (speeds of movement and rotation of the crucible and crystal; design of the heat assembly; flow rate and residual pressure of an inert gas, etc.), concentration main dopant and purity of the feedstock. In this case, the total concentration of germanium and / or tin and / or lead to add them to the silicon melt (charge) is selected from the range of concentrations with respect to silicon (1.0 • 10 ^-7 -15 wt.%):
With _additives = C _germanium + C _tin + C _lead = (1.0 • 10 ^-7 -15 wt.%).

 The introduction of these isovalent additives according to the proposed method allows to increase the stability of twin boundaries and to grow silicon crystals with a cyclic twin structure with a length of more than 300 mm without polycrystalline inclusions.

 In addition, the introduction of germanium, tin, and lead into silicon crystals leads to a change in the band structure of silicon and, to a certain extent, to an expansion of the optical absorption spectrum and a decrease in the generation energy of electron – hole pairs upon irradiation of such crystals with light.

 This leads to an increase in efficiency. solar cells made on the basis of silicon with isovalent additives.

Case Studies
Silicon crystals with a cyclic twin structure with additives of isovalent impurities and without additives were grown by the Czochralski method in a setup for growing Redmet-30 crystals from a crucible with a diameter of 330 mm and loading 22 kg into the crucible. The parameters of all conducted experimental cultivation processes were approximately the same. As the feedstock used was a hole-type conductive silicon with an average specific electrical resistance in the range of 0.9-1.1 Ohm • cm. Cultivation was carried out in an argon flow of 1600 l / h and a residual argon pressure of 10 mm Hg. The crystals were grown with drawing speeds from 1.2 mm / min at the beginning of the cylindrical part to 0.6 mm / min at the end of the growing process. The crystal and crucible rotational speeds were 12 and 5 rpm, respectively. The grown crystals had a diameter in the range of 135-138 mm, and the length of their cylindrical part was 540-570 mm.

 As seed crystals, specially prepared crystals with a given cyclic twin structure oriented in the <110> direction were used. The seed crystals used were rectangular with a cross section of 10 x 10 mm and a length of 120 mm. The seed crystals had a cyclic twin structure of three different types.

 The seed crystal 1 had a basic cyclic twin structure formed by radially spaced two first-order coherent twin planes {111} - {111} and one second-order twin boundary {221} - {221} parallel to the growing direction <110> and intersecting in the center of the seed a crystal. In addition, the seed crystal 1 contained an even number of additional coherent twin planes of the first order parallel to the main coherent twin planes of the basic twin structure.

 The seed crystal 2 had a basic cyclic twin structure (described above) with additional radially arranged parallel and non-parallel coherent twin planes of the first order and three boundaries of the second order twin.

 The seed crystal 3 had a complete cyclic twin structure, including a basic cyclic twin structure with additional radially arranged parallel and non-parallel coherent twin planes of the first order, four twin boundaries of the second order and six twin boundaries of the third order.

A total of 12 crystals were grown. Three crystals were grown without additives, and nine crystals were grown with additives. The stability of the process of growth of silicon crystals with a given cyclic twin structure was evaluated by the critical value of the crystal length (L _kp ), with which the polycrystalline structure began to form. Moreover, the greater the critical length, the higher the stability of the growth process. The results of the growing processes are shown in the table.

 From the results given in the table, it is seen that the stability of the growth process during the growth of silicon crystals with a cyclic twin structure under conditions of doping with germanium, tin and lead additives is significantly higher than in the case of their growth without additives.

Using the proposed method for producing silicon crystals with a cyclic twin structure in comparison with known methods provides the following advantages:
1. The possibility of a significant increase in the length of the grown silicon crystals with a cyclic twin structure without polycrystalline inclusions and, therefore, a significant increase in the productivity of the technology for producing these crystals.

 2. Expansion of the optical absorption spectrum for silicon upon the introduction of isovalent additives with large atomic radii, a decrease in the generation energy of electron-hole pairs upon irradiation with light, and an increase in the efficiency due to this solar cells made on the basis of such a material.

 3. Reducing the cost of silicon crystals with a cyclic twin structure and solar cells made on the basis of these crystals.

Claims (1)

A method for producing silicon crystals with a cyclic twin structure by growing from a melt by the Czochralski method for seed with a cyclic twin structure from a basic three-grain structure formed by two coherent twin planes of the first order and one twin boundary of the second order to a complete cyclic twin structure formed by twenty coherent twin planes first order, with four boundaries of twinning of the second order, with six boundaries of twinning t its order and an even number of additional first-order twinning planes parallel to the specified twenty, characterized in that the cultivation is carried out with the introduction of additives selected from a number of germanium, tin, and lead into the silicon melt, while the concentration of the additives with respect to silicon is 1.0 • 10 ^-7 -15 wt.%.

Images