Synthesis and Purification of Bulk Semiconductors

The synthesis and purification of bulk polycrystalline semiconductor material represents the first step towards the commercial fabrication of an electronic device. This polycrystalline material is then used as the raw material for the formation of single crystal material that is processed to semiconductor wafers. The strong influence on the electric characteristics of a semiconductors exhibited by small amounts of some impurities requires that the bulk raw material be of very high purity (> 99.9999%). Although some level of purification is possible during the crystallization process it is important to use as high a purity starting material as possible. While a wide range of substrate materials are available from commercial vendors, silicon and GaAs represent the only large-scale commercial semiconductor substrates, and thus the discussion will be limited to the synthesis and purification of these materials.

Silicon

Following oxygen (46%), silicon (L. silicis flint) is the most abundant element in the earth's crust (28%). However, silicon does not occur in its elemental form, but as its oxide (SiO2) or as silicates. Sand, quartz, amethyst, agate, flint, and opal are some of the forms in which the oxide appears. Granite, hornblende, asbestos, feldspar, clay and mica, etc. are a few of the numerous silicate minerals. With such boundless supplies of the raw material, the costs associated with the production of bulk silicon is not one of abstraction and conversion of the oxide(s), but of purification of the crude elemental silicon. While 98% elemental silicon, known as metallurgical-grade silicon (MGS), is readily produced on a large scale, the requirements of extreme purity for electronic device fabrication require additional purification steps in order to produce electronic-grade silicon (EGS). Electronic-grade silicon is also known as semiconductor-grade silicon (SGS). In order for the purity levels to be acceptable for subsequent crystal growth and device fabrication, EGS must have carbon and oxygen impurity levels less than a few parts per million (ppm), and metal impurities at the parts per billion (ppb) range or lower. Table 1 and Table 2 give typical impurity concentrations in MGS and EGS, respectively. Besides the purity, the production cost and the specifications must meet the industry desires.

Table 1: Typical impurity concentrations found in metallurgical-grade silicon (MGS).



Element

Concentration (ppm)

Element

Concentration (ppm)


aluminum

1000-4350

manganese

50-120


boron

40-60

molybdenum

< 20 calcium 245-500 nickel 10-105 chromium 50-200 phosphorus 20-50 copper 15-45 titanium 140-300 iron 1550-6500 vanadium 50-250 magnesium 10-50 zirconium 20 Table 2: Typical impurity concentrations found in electronic-grade silicon (EGS). Element Concentration (ppb) Element Concentration (ppb) arsenic < 0.001 gold < 0.00001 antimony < 0.001 iron 0.1-1.0 boron ≤ 0.1 nickel 0.1-0.5 carbon 100-1000 oxygen 100-400 chromium < 0.01 phosphorus ≤ 0.3 cobalt 0.001 silver 0.001 copper 0.1 zinc < 0.1 Metallurgical-grade silicon (MGS) The typical source material for commercial production of elemental silicon is quartzite gravel; a relatively pure form of sand (SiO2). The first step in the synthesis of silicon is the melting and reduction of the silica in a submerged-electrode arc furnace. An example of which is shown schematically in Figure 1, along with the appropriate chemical reactions. A mixture of quartzite gravel and carbon are heated to high temperatures (ca. 1800 °C) in the furnace. The carbon bed consists of a mixture of coal, coke, and wood chips. The latter providing the necessary porosity such that the gases created during the reaction (SiO and CO) are able to flow through the bed. Figure 1: Schematic of submerged-electrode arc furnace for the production of metallurgical-grade silicon (MGS). The overall reduction reaction of SiO2 is expressed in Equation 1, however, the reaction sequence is more complex than this overall reaction implies, and involves the formation of SiC and SiO intermediates. The initial reaction between molten SiO2 and C (Equation 2) takes place in the arc between adjacent electrodes, where the local temperature can exceed 2000 °C. The SiO and CO thus generated flow to cooler zones in the furnace where SiC is formed (Equation 3), or higher in the bed where they reform SiO2 and C (Equation 2). The SiC reacts with molten SiO2 (Equation 4) producing the desired silicon along with SiO and CO. The molten silicon formed is drawn-off from the furnace and solidified. (1) (2) (3) (4) The as-produced MGS is approximately 98-99% pure, with the major impurities being aluminum and iron (Table 1), however, obtaining low levels of boron impurities is of particular importance, because it is difficult to remove and serves as a dopant for silicon. The drawbacks of the above process are that it is energy and raw material intensive. It is estimated that the production of one metric ton (1,000 kg) of MGS requires 2500-2700 kg quartzite, 600 kg charcoal, 600-700 kg coal or coke, 300-500 kg wood chips, and 500,000 kWh of electric power. Currently, approximately 500,000 metric tons of MGS are produced per year, worldwide. Most of the production (ca. 70%) is used for metallurgical applications (e.g., aluminum-silicon alloys are commonly used for automotive engine blocks) from whence its name is derived. Applications in a variety of chemical products such as silicone resins account for about 30%, and only 1% or less of the total production of MGS is used in the manufacturing of high-purity EGS for the electronics industry. The current worldwide consumption of EGS is approximately 5 x 10 6 kg per year. Electronic-grade silicon (EGS) Electronic-grade silicon (EGS) is a polycrystalline material of exceptionally high purity and is the raw material for the growth of single-crystal silicon. EGS is one of the purest materials commonly available, see Table 2. The formation of EGS from MGS is accomplished through chemical purification processes. The basic concept of which involves the conversion of MGS to a volatile silicon compound, which is purified by distillation, and subsequently decomposed to re-form elemental silicon of higher purity (i.e., EGS). Irrespective of the purification route employed, the first step is physical pulverization of MGS followed by its conversion to the volatile silicon compounds. A number of compounds, such as monosilane (SiH4), dichlorosilane (SiH2Cl2), trichlorosilane (SiHCl3), and silicon tetrachloride (SiCl4), have been considered as chemical intermediates. Among these, SiHCl3 has been used predominantly as the intermediate compound for subsequent EGS formation, although SiH4 is used to a lesser extent. Silicon tetrachloride and its lower chlorinated derivatives are used for the chemical vapor deposition (CVD) growth of Si and SiO2. The boiling points of silane and its chlorinated products (Table 3) are such that they are conveniently separated from each other by fractional distillation. Table 3: Boiling points of silane and chlorosilanes at 760 mmHg (1 atmosphere). Compound Boiling point (°C) SiH4 -112.3 SiH3Cl -30.4 SiH2Cl2 8.3 SiHCl3 31.5 SiCl4 57.6 The reasons for the predominant use of SiHCl3 in the synthesis of EGS are as follows: 1.SiHCl3 can be easily formed by the reaction of anhydrous hydrogen chloride with MGS at reasonably low temperatures (200 - 400 °C); 2.it is liquid at room temperature so that purification can be accomplished using standard distillation techniques; 3.it is easily handled and if dry can be stored in carbon steel tanks; 4.its liquid is easily vaporized and, when mixed with hydrogen it can be transported in steel lines without corrosion; 5.it can be reduced at atmospheric pressure in the presence of hydrogen; 6.its deposition can take place on heated silicon, thus eliminating contact with any foreign surfaces that may contaminate the resulting silicon; and 7.it reacts at lower temperatures (1000 - 1200 °C) and at faster rates than does SiCl4. Chlorosilane (Seimens) process Trichlorosilane is synthesized by heating powdered MGS with anhydrous hydrogen chloride (HCl) at around 300 °C in a fluidized-bed reactor, Equation 5. (5) Since the reaction is actually an equilibrium and the formation of SiHCl3 highly exothermic, efficient removal of generated heat is essential to assure a maximum yield of SiHCl3. While the stoichiometric reaction is that shown in Eq. 5, a mixture of chlorinated silanes is actually prepared which must be separated by fractional distillation, along with the chlorides of any impurities. In particular iron, aluminum, and boron are removed as FeCl3 (b.p. = 316 °C), AlCl3 (m.p. = 190 °C subl.), and BCl3 (b.p. = 12.65 °C), respectively. Fractional distillation of SiHCl3 from these impurity halides result in greatly increased purity with a concentration of electrically active impurities of less than 1 ppb. EGS is prepared from purified SiHCl3 in a chemical vapor deposition (CVD) process similar to the epitaxial growth of Si. The high-purity SiHCl3 is vaporized, diluted with high-purity hydrogen, and introduced into the Seimens deposition reactor, shown schematically in Figure 2. Within the reactor, thin silicon rods called slim rods (ca. 4 mm diameter) are supported by graphite electrodes. Resistance heating of the slim rods causes the decomposition of the SiHCl3 to yield silicon, as described by the reverse reaction shown in Eq. 5. Figure 2: Schematic representation of a Seimens deposition reactor. The shift in the equilibrium from forming SiHCl3 from Si at low temperature, to forming Si from SiHCl3 at high temperature is as a consequence of the temperature dependence (Equation 6) of the equilibrium constant (Equation 7, where ρ = partial pressure) for Equation 5. Since the formation of SiHCl3 is exothermic, i.e., ΔH < 0, an increase in the temperature causes the partial pressure of SiHCl3 to decrease. Thus, the Siemens process is typically run at ca. 1100 °C, while the reverse fluidized bed process is carried out at 300 °C. (6) (7) The slim rods act as a nucleation point for the deposition of silicon, and the resulting polycrystalline rod consists of columnar grains of silicon (polysilicon) grown perpendicular to the rod axis. Growth occurs at less than 1 mm per hour, and after deposition for 200 to 300 hours high-purity (EGS) polysilicon rods of 150-200 mm in diameter are produced. For subsequent float-zone refining the polysilicon EGS rods are cut into long cylindrical rods. Alternatively, the as-formed polysilicon rods are broken into chunks for single crystal growth processes, for example Czochralski melt growth. In addition to the formation of silicon, the HCl coproduct reacts with the SiHCl3 reactant to form silicon tetrachloride (SiCl4) and hydrogen as major byproducts of the process, Equation 8. This reaction represents a major disadvantage with the Seimens process: poor efficiency of silicon and chlorine consumption. Typically, only 30% of the silicon introduced into CVD reactor is converted into high-purity polysilicon. (8) In order to improve efficiency the HCl, SiCl4, H2, and unreacted SiHCl3 are separated and recovered for recycling. Figure 3 illustrates the entire chlorosilane process starting with MGS and including the recycling of the reaction byproducts to achieve high overall process efficiency. As a consequence, the production cost of high-purity EGS depends on the commercial usefulness of the byproduct, SiCl4. Additional disadvantages of the Seimens process are derived from its relatively small batch size, slow growth rate, and high power consumption. These issues have lead to the investigation of alternative cost efficient routes to EGS. Figure 3: Schematic representation of the reaction pathways for the formation of EGS using the chlorosilane process. Silane process An alternative process for the production of EGS that has begun to receive commercial attention is the pyrolysis of silane (SiH4). The advantages of producing EGS from SiH4 instead of SiHCl3 are potentially lower costs associated with lower reaction temperatures, and less harmful byproducts. Silane decomposes < 900 °C to give silicon and hydrogen, Equation 9. (9) Silane may be prepared by a number of routes, each having advantages with respect to purity and production cost. The simplest process involves the direct reaction of MGS powders with magnesium at 500 °C in a hydrogen atmosphere, to form magnesium silicide (Mg2Si). The magnesium silicide is then reacted with ammonium chloride in liquid ammonia below 0 °C, Equation 10. (10) This process is ideally suited to the removal of boron impurities (a p-type dopant in Si), because the diborane (B2H6) produced during the reaction forms the Lewis acid-base complex, H3B(NH3), whose volatility is sufficiently lower than SiH4, allowing for the purification of the latter. It is possible to prepare EGS with a boron content of ≤ 20 ppt using SiH4 synthesized in this manner. However, phosphorus (another dopant) in the form of PH3 may be present as a contaminant requiring subsequent purification of the SiH4. Alternative routes to SiH4 involve the chemical reduction of SiCl4 by either lithium hydride (Equation 11), lithium aluminum hydride (Equation 12), or via hydrogenation in the presence of elemental silicon (Equation 13 - Equation 16). The hydride reduction reactions may be carried-out on relatively large scales (ca. 50 kg), but only batch processes. In contrast, Union Carbide has adapted the hydrogenation to a continuous process, involving disproportionation reactions of chlorosilanes (Equation 14 - Equation 16) and the fractional distillation of silane (Table 3). (11) (12) (13) (14) (15) (16) Pyrolysis of silane on resistively heated polysilicon filaments at 700-800 °C yields polycrystalline EGS. As noted above, the EGS formed has remarkably low boron impurities compared with material prepared from trichlorosilane. Moreover, the resulting EGS is less contaminated with transition metals from the reactor container because SiH4 decomposition does not cause as much of a corrosion problem as is observed with halide precursor compounds. Granular polysilicon deposition Both the chlorosilane (Seimens) and silane processes result in the formation of rods of EGS. However, there has been increased interest in the formation of granular polycrystalline EGS. This process was developed in 1980’s, and relies on the decomposition of SiH4 in a fluidized-bed deposition reactor to produce free-flowing granular polysilicon. Tiny silicon particles are fluidized in a SiH4/H2 flow, and act as seed crystal onto which polysilicon deposits to form free-flowing spherical particles. The size distribution of the particles thus formed is over the range from 0.1 to 1.5 mm in diameter with an average particle size of 0.7 mm. The fluidized-bed seed particles are originally made by grinding EGS in a ball (or hammer) mill and leaching the product with acid, hydrogen peroxide, and water. This process is time-consuming and costly, and tended to introduce undesirable impurities from the metal grinders. In a new method, large EGS particles are fired at each other by a high-speed stream of inert gas and the collision breaks them down into particles of suitable size for a fluidized bed. This process has the main advantage that it introduces no foreign materials and requires no leaching or other post purification. The fluidized-bed reactors are much more efficient than traditional rod reactors as a consequence of the greater surface area available during CVD growth of silicon. It has been suggested that fluidized-bed reactors require 1/5 to 1/10 the energy, and half the capital cost of the traditional process. The quality of fluidized-bed polysilicon has proven to be equivalent to polysilicon produced by the conventional methods. Moreover, granular EGS in a free-flowing form, and with high bulk density, enables crystal growers to obtain the high, reproducible production yields out of each crystal growth run. For example, in the Czochralski crystal growth process, crucibles can be quickly and easily filled to uniform loading with granular EGS, which typically exceed those of randomly stacked polysilicon chunks produced by the Siemens silane process. Zone refining The technique of zone refining is used to purify solid materials and is commonly employed in metallurgical refining. In the case of silicon may be used to obtain the desired ultimate purity of EGS, which has already been purified by chemical processes. Zone refining was invented by Pfann, and makes use of the fact that the equilibrium solubility of any impurity (e.g., Al) is different in the solid and liquid phases of a material (e.g., Si). For the dilute solutions, as is observed in EGS silicon, an equilibrium segregation coefficient (k0) is defined by k0 = Cs/Cl, where Cs and Cl are the equilibrium concentrations of the impurity in the solid and liquid near the interface, respectively. If k0 is less than 1 then the impurities are left in the melt as the molten zone is moved along the material. In a practical sense a molten zone is established in a solid rod. The zone is then moved along the rod from left to right. If k < 1 then the frozen part left on the trailing edge of the moving molten zone will be purer than the material that melts in on the right-side leading edge of the moving molten zone. Consequently the solid to the left of the molten zone is purer than the solid on the right. At the completion of the first pass the impurities become concentrated to the right of the solid sample. Repetition of the process allows for purification to exceptionally high levels. Table 4. lists the equilibrium segregation coefficients for common impurity and dopant elements in silicon; it should be noted that they are all less than 1. Table 4: Segregation coefficients for common impurity and dopant elements in silicon. Element k0 Element k0 aluminum 0.002 iron 8 x 10-6 boron 0.8 oxygen 0.25 carbon 0.07 phosphorus 0.35 copper 4 x 10-6 antimony 0.023 Gallium arsenide In contrast to electronic grade silicon (EGS), whose use is a minor fraction of the global production of elemental silicon, gallium arsenide (GaAs) is produced exclusively for use in the semiconductor industry. However, arsenic and its compounds have significant commercial applications. The main use of elemental arsenic is in alloys of Pb, and to a lesser extent Cu, while arsenic compounds are widely used in pesticides and wood preservatives and the production of bottle glass. Thus, the electronics industry represents a minor user of arsenic. In contrast, although gallium has minor uses as a high-temperature liquid seal, manometric fluids and heat transfer media, and for low temperature solders, its main use is in semiconductor technology. Isolation and purification of gallium metal At 19 ppm gallium (L. Gallia, France) is about as abundant as nitrogen, lithium and lead; it is twice as abundant as boron (9 ppm), but is more difficult to extract due to the lack of any major gallium-containing ore. Gallium always occurs in association either with zinc or germanium, its neighbors in the periodic table, or with aluminum in the same group. Thus, the highest concentrations (0.1-1%) are in the rare mineral germanite (a complex sulfide of Zn, Cu, Ge, and As), while concentrations in sphalerite (ZnS), diaspore [AlO(OH)], bauxite, or coal, are a hundred-fold less. Industrially, gallium was originally recovered from the flue dust emitted during sulfide roasting or coal burning (up to 1.5% Ga), however, it is now obtained as side product of vast aluminum industry and in particular from the Bayer process for obtaining alumina from bauxite. The Bayer process involves dissolution of bauxite, AlOxOH3-2x, in aqueous NaOH, separation of insoluble impurities, partial precipitation of the trihydrate, Al(OH)3, and calcination at 1,200 °C. During processing the alkaline solution is gradually enriched in gallium from an initial weight ratio Ga/Al of about 1/5000 to about 1/300. Electrolysis of these extracts with a Hg cathode results in further concentration, and the solution of sodium gallate thus formed is then electrolyzed with a stainless steel cathode to give Ga metal. Since bauxite contains 0.003-0.01% gallium, complete recovery would yield some 500-1000 tons per annum, however present consumption is only 0.1% of this about 10 tons per annum. A typical analysis of the 98-99% pure gallium obtained as a side product from the Bayer process is shown in Table 5. This material is further purified to 99.99% by chemical treatment with acids and O2 at high temperatures followed by crystallization. This chemical process results in the reduction of the majority of metal impurities at the ppm level, see Table 5. Purification to seven nines 99.9999% is possible through zone refining, however, since the equilibrium distribution coefficient of the residual impurities k0 ≈ 1, multiple passes are required, typically > 500. The low melting point of gallium ensures that contamination from the container wall (which is significant in silicon zone refining) is minimized. In order to facilitate the multiple zone refining in a suitable time, a simple modification of zone refining is employed shown in Figure 4. The gallium is contained in a plastic tube wrapped around a rotating cylinder that is half immersed in a cooling bath. A heater is positioned above the gallium plastic coil. Thus, establishing a series of molten zones that pass upon rotation of the drum by one helical segment per revolution. In this manner, 500 passes may be made in relatively short time periods. The typical impurity levels of gallium zone refined in this manner are given in Table 5.

Table 5: Typical analysis of gallium obtained as a side product from the Bayer process.



Element

Bayer process (ppm)

After acid/base leaching (ppm)

500 zone passes (ppm)


aluminum

100-1,000

7

< 1 calcium 10-100 not detected not detected copper 100-1,000 2 < 1 iron 100-1,000 7 < 1 lead < 2000 30 not detected magnesium 10-100 1 not detected mercury 10-100 not detected not detected nickel 10-100 not detected not detected silicon 10-100 ≈ 1 not detected tin 10-100 ≈ 1 not detected titanium 10-100 1 < 1 zinc 30,000 ≈ 1 not detected Figure 4: Schematic representation of a zone refining apparatus. Isolation and purification of elemental arsenic Elemental arsenic (L. arsenicum, yellow orpiment) exists in two forms: yellow (cubic, As4) and gray or metallic (rhombohedral). At a natural abundance of 1.8 ppm arsenic is relatively rare, however, this is offset by its presence in a number of common minerals and the relative ease of isolation. Arsenic containing minerals are grouped into three main classes: the sulfides realgar (As4S4) and orpiment (As2S3), the oxide arsenolite (As2O3), and the arsenides and sulfaresenides of the iron, cobalt, and nickel. Minerals in this latter class include: loellinginite (FeAs2), safforlite (CoAs), niccolite (NiAs), rammelsbergite (NiAs2), ansenopyrite or mispickel (FeAsS), cobaltite (CoAsS), enargite (Cu3AsS4), gerdsorfite (NiAsS), and the quarturnary sulfide glaucodot [(Co,Fe)AsS]. Table 6 shows the typical impurities in arsenopyrite. Table 6: Typical impurities in arsenopyrite. Element Concentration (ppm) Element Concentration (ppm) silver 90 nickel < 3,000 gold 8 lead 50 cobalt 30,000 platinum 0.4 copper 200 rhenium 50 germanium 30 selenium 50 manganese 3,000 vanadium 300 molybdenum 60 zinc 400 Arsenic is obtained commercially by smelting either FeAs2 or FeAsS at 650-700 °C in the absence of air and condensing the sublimed element (Tsub = 613 °C), Equation 17. (17) The arsenic thus obtained is combined with lead and then sublimed (Tsub = 614 °C) which binds any sulfur impurities more strongly than arsenic. Any residual arsenic that remains trapped in the iron sulfide is separated by forming the oxide (As2O3) by roasting the sulfide in air. The oxide is sublimed into the flue system during roasting from where it is collected and reduced with charcoal at 700-800 °C to give elemental arsenic. Semiconductor grade arsenic (> 99.9999%) is formed by zone refining.

Synthesis and purification of gallium arsenide.

Gallium arsenide can be prepared by the direct reaction of the elements, Equation 18. However, while conceptually simple the synthesis of GaAs is complicated by the different vapor pressures of the reagents and the highly exothermic nature of the reaction. Furthermore, since the synthesis of GaAs at atmospheric pressure is accompanied by its simultaneous decomposes due to the loss by sublimation, of arsenic, the synthesis must be carried out under an overpressure of arsenic in order to maintain a stoichiometric composition of the synthesized GaAs.
(18)

In order to overcome the problems associated with arsenic loss, the reaction is usually carried out in a sealed reaction tube. However, if a stoichiometric quantity of arsenic is used in the reaction a constant temperature of 1238 °C must be employed in order to maintain the desired arsenic overpressure of 1 atm. Practically, it is easier to use a large excess of arsenic heated to a lower temperature. In this situation the pressure in the tube is approximately equal to the equilibrium vapor pressure of the volatile component (arsenic) at the lower temperature. Thus, an over pressure of 1 atm arsenic may be maintained if within a sealed tube elemental arsenic is heated to 600-620 °C while the GaAs is maintained at 1240-1250 °C.
Figure 5 shows the sealed tube configuration that is typically used for the synthesis of GaAs. The tube is heated within a two-zone furnace. The boats holding the reactants are usually made of quartz, however, graphite is also used since the latter has a closer thermal expansion match to the GaAs product. If higher purity is required then pyrolytic boron nitride (PBN) is used. One of the boats is loaded with pure gallium the other with arsenic. A plug of quartz wool may be placed between the boats to act as a diffuser. The tube is then evacuated and sealed. Once brought to the correct reaction temperatures (Figure 5), the arsenic vapor is transported to the gallium, and they react to form GaAs in a controlled manner. Table 7 gives the typical impurity concentrations found in polycrystalline GaAs.

Figure 5: Schematic representation of a sealed tube synthesis of GaAs.






Table 7: Impurity concentrations found in polycrystalline GaAs.



Element

Concentration (ppm)

Element

Concentration (ppm)


boron

0.1

silicon

0.02


carbon

0.7

phosphorus

0.1


nitrogen

0.1

sulfur

0.01


oxygen

0.5

chlorine

0.08


fluorine

0.2

nickel

0.04


magnesium

0.02

copper

0.01


aluminum

0.02

zinc

0.05

Polycrystalline GaAs, formed in from the direct reaction of the elements is often used as the starting material for single crystal growth via Bridgeman or Czochralski crystal growth. It is also possible to prepare single crystals of GaAs directly from the elements using in-situ, or direct, compounding within a high-pressure liquid encapsulated Czochralski (HPLEC) technique.