Polycrystalline Silicon Study Low-Cost Silicon Refining Technology Prospects and Semiconductor-Grade Polycrystalline Silicon Availability through 1 988

Photovoltaic arrays that convert solar energy into electrical energy can become a cost-effective bulk energy generation alternative, provided that an adequate supply of low-cost materials is available. One of the key requirements for economic photovoltaic cells is reasonably priced silicon. At present, the photovoltaic industry is dependent upon polycrystalline silicon refined by the Siemens process primarily for integrated circuits, power deviczs, and discrete semiconductor devices. Tnis dependency is expected to continue until the DOE-sponsored low-cost silicon refining technology developments have matured to the point where they are in commercial use. The photovoltaic industry can then develop its own source of supply. Since 1979 the Jet Propulsion Laboratory Technology Development and Application Lead Center (TD&A), now the Program Analysis and Integration Center (PA&I), has periodically examined the availability of refined silicon and the status of the DOE-sponsored low-cost silicon refining technology developments. Tnree reports have been published, based on surveys conducted by JPL consultants in 1979, 1.981, and 1983. This report updates the silicon material availability and market pricing projections through 1988, based on data collected early in 1984. It also overviews the silicon refining industry plans to meet the increasing demands of the semiconductor device and photovoltaic product industries. A section of the report has been devoted to review of the DOE-sponsored technology research for producing low-cost polycrys,talline silicon, probabilistic cost analysis for the two most promising production processes for achieving the DOE cost goals, and the impacts of the DOE photovoltaics program silicon refining research upon the commercial polycrystalline silicon refining industry.

Since 1979 the Jet Propulsion Laboratory (JPL) Photovoltaics Program Analysis and Integration Center (PA&I) has periodically evaluated the availability of refined silicon and the status of new technology developments. Silicon material outlook studies have been previously published to report on the status of the silicon refining industry and to predict silicon availability. In November 1979 a survey (~eference 1) concluded that a shortage of silicon material was developing because of steadily increasing consumption by the semiconductor products industry with no concomitant increase in production capacity. Such a shortage would have hurt the growth of the PV industry because the selling price of any available silicon material would have been too high for the PV industry to use. 'Ihe published report was widely accepted by the silicon refining industry. It provided for them, for the first time, a complete review of the silicon material supply-and-demand situation, includir~g data for the present and a forecast of the demand of the PV industry.
An apparent reason for the developing shortage was that the silicon refining industry had been concerned that the DOE-funded program to develop technology for low-cost polycrystalline production would lead to processes with substantial cost advantages, and thus production by the Siemens process would beome less competitive. This concern led to a reluctance by the industry to invest in construction of new plants using the Siemens process. As the number of new technology options was narrowed to the two most promising contenders, the Z i o n Carbide Corp. (UCC) silane process and the Hemlock Semiconductor Corp. modified Siemens process, the refining industry's concern for outmoding of the Siemens process abated.
The silicon industry is highly competitive, and is striving to reduce production cost. Intensive proprietary efforts were made to develop improvements in the Siemens process. As a cesult, modifications of the Siemens reactor and of the process for the preparation and purification of trichlorosilane have been installed in existing plants. These have led to decreased production costs, especially in electric energy use for deposition reactors. Throughput and material utilization efficiency have increased greatly through these changes.
In November 1980 a survey of the silicon refining industry (Reference 2) conducted by Remo Pellin, a JPL consultant, found that silicon availability had begun to change; however, this was not before some large fluctuations in the spot-market price had taken place. The spot-market price for sil.icon was quoted as high as $140/kg in April 1980; the lowest 1980 spot-market price quoted was below $50/kg.
Another update of the silicon availability analysis, conducted in December 1982 and January 1983 (Reference 31, indicated that the improved trend of adequate silicon supply for the photovoltaic and semiconductor industries was continuing. The conclusions of this survey were:

1)
Semiconductor-grade polycrystalline silicon will be available to meet the market demands of the semiconductor industry although some transient situations of short supply may occur, depending on the rate of industry recovery from the recession experienced in 1980.
2) Lower-grade polycrystalline silicon (rejected by the semiconductor industry) would be available for the PV industry at the market price. This price might: be more than the industry could afford to pay while continuing to reduce the price of the PV products in its strategy to accelerate market demands.
The subjects addressed in this report are: (1) a final survey of silicon material availability; the situation is considered to be well in hand, with industry aggressively maintaining supply near the demands; (2) technology status and expansion plans for the silicon refining industry; ( 3 ) a brief look at the polysilicon market pricing outlook, particularly for Lhe photovoltaic industry, and (4) a status review of research funded by DOE for the development of low-cost processes for polysilicon production, and the effect uf that research on the polysilicon industry.
The industry survey data were collected primarily by Remo Pellin, with assistance by James Lorenz, another JPL consultant. Additional input was provided and analyses were conducted by PA&I and Flat-Plate Solar Array Project (FsA) staff members.
The conclusions from the data collected are: (1) in the past, the demands of the semiconductor industry and the PV industry have been met by production capacity increases; additional production capacity increases are planned to meet the demand through 1388. Table 1 shows the totals of world capacity to manufacture semiconductor-grade polycrystalline silicon and of silicon usage. Although the data in Table 1 showed that the forecast plant, capacity may be in excess of forecast demand, the expected production rate will be closely matched to the market demands, as has been the case in the past; (2) the silicon refining industry, as in the past, will continue to improve the Siemens refining process to reduce cost and improve product quality, and the entry of UCC into the polysilicon business will promote strong competition; (3) the market price of silicon will depend on the supply-and-demand situation, with sporadic high spot-market prices. The demand will be sensitive to growth rates of both the semiconductor and the PV industries. The DOE-sponsored research program has had significant effects on the world's polysilicon production technology and production capacity. Tne decision for the commercialization of the silane process by UCC is a prime example of Government-sponsored, high-risk research yielding a promising result with rapid transfer to industry. UCC had considered entering the polysilicon business several times before 1980, but did not make the commitment until the feasibility of the silane process had been demonstrated in the DOE-funded, FSA-administered research program.
Today, and in the near future, most polycrystalline silicon will be manufactured by variants of the Siemens process. The modified Siemens process elements, developed by Hemlock with DOE funding, can ultimately reduce the production cost of all polycrystalline silicon produced by Siemens-type processes. Lower production cost in the modified Siemens processes and in commercialized low-cost processes is expected to influence the market price, which is expected to decrease gradually, depending on the supply/demand scenario. It can be said that the DOE-sponsored low-cost silicon refining research is one area in which the United States, through proper use of Government resources, has maintained technology leadership. FSA silicon refining research has initiated industry action and focused attention on this critical polysilicon technology.

INTRODUCTZON
The practical.ity of photovoltaic (PV) arrays as an energy source depends upon its energy selling price compared with that of other energy sources. Because semiconductor-grade polycrystalline silicon (Si) is almost universally used as the prime material for the manufacture of photovoltaic arrays, the availability of a large quantity of relatively low-cost semiconductor-grade polycrystalline gilicon is critical.
At present, the photovoltaics industry depends upon the availability of semiconductor-grade polycrystalline silicon, which is produced for the semicorlductor industry and is used primarily for integrated circuits, power devices, and discrete semi.conductor devices. This is expected to be the primary source until DOE-sponsored research provides the technology for low-cost processes and the refining industry installs production capacity based on this technology.
The Jet Propulsion 1,aboratory's (JPL's) Photovoltaics Program Analysis and Integration Center (PAEI) has tracked the availability of the silicon material (critical for development of the photovoltaics industry) and has periodically reported survey results. (1) A severe shortage could occur in 1980 due to increased consumption by the semiconductor products industry and by the emerging photovoltaics industry. Very little expansion of the production capacity using existing technology was expected during this period because of the prospect of the new low-cost refining technology under development by the DOE photovoltaics program.
( 2 ) As a result, a seller's market would be created and the price af silicon would increase.
In November 1980 (Reference 2), a silicon industry survey update was conducted by a JPL consultant, Remo Pellin. It was concluded that the silicon market outlook was beginning to change significantly. However, before that, some large fluctuations in spot-market pricing had taken place. In April 1980, thc spot-market price for silicon was quoted as high as $140/kg (Reference 3 1 , ail increase of a factor of two from the spot-market prices quoted in 1979. The survey data were summarized in a report titled Industry Survey Report Status of Silicon Material and Silicon Sheet Techniques (~eference 21, with the following conclusions : (1) No long-term shortage of semiconductor-grade polycrystalline silicon would occur and the needs of the semiconductor products industry would be met. This improved supply outlook was the r e s p o n s e o f t h e p o l y s i l i c o n r e f i n i n g i n d u s t r y t o t h e f o r e c a u t s h o r t f a l l . The i n d u s t r y had begun t o i n c r e a s e p l a n t c a p a c i t y i n 1980, p r i m a r i l y by p r o c e s s improvements o f e x i s t i n g p l a n t s .
( ( 3 ) G r e a t l y expanded p l a n t c a p a c i t y h a s been a c h i e v e d w i t h minimal c a p i t a l e x p e n d i t u r e s through p r o c e s s improvements.

SILICON MARKETPLACE
A.
POLYCRY STALLINE SILICON OUTLOOK Silicon solar cells are manufactured from single-crystal wafezs, polycrystalline wafers, ribbons, or amorphous silicon thin films. The cost of the polycrystalline silicon currently constitutes about 10% to 15% of the manufacturing cost of a typical photovoltaic module ($0.60 to $0.90 out of $6 per peak watt). The use of semiconductor-grade p~lycryst.~~lline silicon allows the manufacture of highly efficient single-crystal solar ct'lls, reducing the overall cost of the module and support structure per peak watt.
It has been concluded that the use of any special "solar-grade" polycrystalline silicon, which results in noticeably less efficient solar cells, is not warranted, as the program emphasis i.s on high efficiency. Thus, the silicon solar-cell industry depends for supply on the same polycrystalline silicon as does the semiconductor industry. For the near future, as in the past, the silicon solar cell industry will be able to obtain and use rejected polycrystalline silicon. This rejected silicon consists of virgin material that does not meet integrated-circuit impurity content specifications for such elements as boron, phosphorus, or carbon; this rejected silicon is as much as 5% of all semiconductor-grade silicon produced. In general, solar cells with satisfactory efficiency can be fabricated from this silicon.
Major changes in the semiconductor-grade polycrystalline silicon market occurred in 1983. In January 1983 the semiconductor product industry began a mild recovery from the 1981-J982 world recession. During J u n~ 1983 the recession suddenly ended in the integrated circuit product industry and panic buying began (~eference 5). This sales boom continued through March 1984 and is expected to extend at least through 1985 (Reference 5).
As a result of this boom in the sales of integrated circuits, orders for polished si,,icon wafers increased by 50% between July 1983 and December 1983.
Only the presence of a significant unused capacity for the production of polished silicon wafers prevented panic buying of this product. The annual world capacity for the manufacture of polished silicon wafers was 1.5 billion square inches in early 1984. The. projected demand is expected to be 1.1 billion square inches in 1984 and 1.385 billion square inches in 1985. Expansion of wafer manufacturing capacity is relatively easy, and new wafer plants can be built in 12 to 15 months. No shortage of poiished wafers is expected through 1990. Indications are that the expected wafer manufacturing increases will take place primarily in four companies, only one of which is i.n the United States: Wacker Chemetronic GmbH, Monsanto Co., Shin-Etsu Handota-, and Osaka Titanium Mfg. Co.
The use of semiconductor-grade polycrystalline silicon has grown commensurately with the increase in use of polished silicon wafers. In March 1984 the po1ycr;~stalline silicon demand and production were balanced (Reference 6). Even though production capacity is expanding, some experts currently believe that a shortage of polycrystalline silicon could occur in to build silicon refining plants. A careful analysis, however, reveals that the production capacity, con,,prising existing plants, plants being built, and plants in the planning stage, should provide sufficient polycrystalline silicon to meet the forecast demand, at least through 1988. Accordingly, it is expected that prices for polycrystalline silicon will slowly decline without large transients over this period because of competitive pressures, large-scale manufacture, and new-technology plants beginning operation.
Various factors in the production and use of polycrystalline silicon require some explanation. Contrary to the normal business cycle, at the beginning of the recession recovery period in the second half of 1983, increases in polycrystalline purchases were larger than expected. It is expected that this increase will soon settle back to a normal rate. Several months are usually required to convert polycrystalline silicon into integrated circuits; business cycles seldom are predictable beyond six months. Ihus, buying requests normally increase at the start of the boom period and slow down at the, end of it. Hence, raw materials are always more difficult to purchase at the start of a boom business cycle. In the usual business ;;ractice, raw material inventories are increased substanti.ally above steady state. These factors virtually ensure that some panic buying of raw materials will occur, and this has occurred for polycrystalline silicon during the past six months. m e capacity of a polycrystalline silicon plant should be considered to be variable. For example, both Wacker and Hemlock Semiconductor Corp. have demonstrated that the capacity of an existing plant can be doubled by process changes alone. Tnis has been done by introducing the reactants at a higher rate, by increasing the average rod diameter of produced polycrystall.ine silicon, and by increasing the deposition temperature. These parametric adjustments can be made to increase capacity of existing plants during times of high demand.
TO predict the status of the p~l~crystalline silicon market in future years, an analysis must be made of both the market demand and manufacturing capacity. Corrections were also necessary to reflect lower yields for the Japanese Shin-Etsu Handotai and Osaka Titanium companies in converting polycrystalline silicon to polished silicon wafers; the Japanese yield is approximately 35% versus 50% for both Wacker and Monsanto. However, the Japanese share of the world market is increasing significantly because of the high wafer quality they supply as a result of their stringent product specifications. Based on Table 2-1 data, polycrystalline silicon use will continue to increase from approximately 4,000 metric tons in 1984 to more than 10,000 metric tons in 1988. Although boom and recession business cycles are difficult to predict and the exact yearly usage may vary from Table 2-1, the overall growth projection of about 20% per yeai is reasonable through 1988.   Table 2-2 records the free-world polycrystalline silicon usage for the years 1977 to 1983. The table shows the industries that convert polycrystalline silicon to polished silicon wafers. Some companies such as Texas Instruments, Inc., Motorola Inc., and IBM manufacture polished wafers f0.r internal use. Others such as Wacker, Monsanto, and Shin-Etsu Handotai manufacture wafers for sale. 1983 was a difficult time for all of these companies. During the first half of the year, sales were hard to generate; during the second half of the year, facility start-up problems had to be addressed to meet the rapidly increasing demand.
A forecast of free-world us< of polycrystalline silicon through 1988 is presented in Table 2-3. The followi.ng wafering-related observations are implicit in this table: (1) No new company will beco:ne a major manufacturer of polished silicon wafers.
(2) Merchant suppliers of polished silicon wafers will take over the business from the captive (device) manufacturers.
The complemeritary metal-oxide semiconductor (CMOS) technology will largely replace n-type metal-oxide semiconductor (NMOS) technology for the manufacture of silicon-integrated circuits during the next five years. This will require that polished-siJ-icon wafer manufacturers add the epitaxy step to their processes and that they sell primarily epitaxial wafers rather than polished-silicon wafers. This change will entail some silicon feedstock losses in the epitaxial process, but will provide some gains in integratedcircuit yields. Overall, it is expected that these two factors will balance, and no net change in the use of polycrystalline siiicon will occur due to this process step addition.
The world capacity for production of semiconductor-grade polycrystalline silicon for the years from 1977 to 1983 is presented in Table 2-4. The rated capacity shown in Table 2-4 and the actual production are two different sets of numbers. Polycrystalline silicon production and usage mmbers also differ substantially. In the past, polycrystalline silicon refining plants frequently ran at capacity, and the product was either sold or stockpiled for future sales. In the future, this stockpiling practice will largely be discontinued because of the high-interest cost of inventory and the lower future market prices expected.
A world-capacity forecast of semiconductor-grade polycrystalline silicon production is made in Table 2-5. The projected growth rate of this capacity is impressive.

B. REFINING INDUSTRY STATUS AND PLANS
The recent major changes in the semiconductor-grade silicon market sales, the technology improvement for silicon refining, and the prospect of increased market demand have forced important decisions and plans for the ref iriing industry that are worth mentioning.    Motorola and Monsanto were planned because of production costs; the planned shutdown of the Texas Instruments plant was apparently based on silicon-purity issues.
The Wacker plant in Burghausen, West Germany, will continue to use the Siemens process with trichlorosilane to manufacture polycrystalline silicon. Wacker believes that the risks involved in the use of silane or dichlorosilane ma1.e ' . t not worthwhile to convert the plant. Additional process changes will allow Wacker to increase the capacity of its Burghausen plant to at least 2500 metric tons. Wacker believes that fluidized-bed continuous-deposition technology using trichlorosilane (TCS) feedstock will not be on stream commercially until after 1990, and this will force other expansion decisions. Wacker improved the cost basis for the manufacture of polycrystailine silicon substantially during 1983 by installing a plant to recycle and hydrogenate the by-product silicon tetrachloride to TCS. With low-cost electrical power, on-site manufacture of TCS, and hydrogenation of by-product silicon tetrachloride, Wacker is today's lowest-cost producer of electronic-grade polycrystalline silicon.
Hemlock Semiconductor Corp. presently has an old, high-cost plant for the manufacture of polycrystalline silicon. The parent company, Dow Corning Corp., has been reluztant to invest large quantities of new money to renovate the plant. Shin-Etsu Handotai has recently purchased a 24-112% interest in Hemlock. Another 24-112% interest is likely to be sold in a joint venture arrangement. As a part of this joint venture, Hemlock is expected to make the necessary capital investments and add process improvements for the plant to become a low-cost producer. Among the changes expected are the recycling and hydrogenation of the procesq'by-product silicon tetrachloride to TCS. The process changes have the potenefal to increase the capacity of the Hemlock plant and reduce production costs significantly.
The Japanese governnent has recognized that one of its huge industries is dependent upon imported polycrystalline silicon and has initiated efforts to produce low-cost silicon in Japan. The New Energy Development Organization (NEDO) has recently completed a TCS-manufacturing facility. This facility consists of a vertical fluidized-bed-type reaction tower filled with fine metallic silicon granules, into which silicon tetrachloride gas is blown from underneath to synthesize TCS. The Osaka Titanium fact,~ry has a production capacity of 200 tonslyr of TCS. It is also anticipated that Osaka Titanium, with joint ventures, will increase its polycrystalline silicon plant capacity to 800 metric tons in 1985 and to 1800 metric tons by 1990. Osaka Titanium has also conducted its own research to refine metallurgical-grade silicon to a product less pure than semiconductor grade. The patents granted indicate that a variant of the Siemens process will be used at Osaka Titanium with on-site silicon tetrachloride by-product recycling and hydrogenation.
NED0 also has developed an energy-efficient silicon refining process through a research and development project at Shin-Etsu Chemical Co., Ltd. In this process a fluidized bed is used to produce a solar-grade silicon continuously, and the unit power requirement (quantity of electricity required to produce 1 kg) is only about 30 kWh. Ibis is much lower than the energy required Both w i l l p r o b a b l y s h u t down a s soon a s a n a d e q u a t e a l t e r n a t i v e s u p p l y of p o l y c r y s t a l l i n e s i l i c o n i s g u a r a n t e e d .
General E l e c t r i c Co. purchased t h e G r e a t Western S i l i c o n Corp. e a r l y i n 1981. This p l a n t u s e s purchased TCS i n t h e Siemens p r o c e s s . GE had n o t i n c r e a s e d t h e p l a n t ' s c a p a c i t y u n t i l r e c e n t l y . The p l a n t w i l l e i t h e r b e expanded t o l a r g e c a p a c i t y w i t h o n -s i t e r e c y c l i n g o r GE w i l l c e a s e o p e r a t i o n of t h e p l a n t .
Shin-.Etsu Handotai and Komatsu E l e c t r o n i c M e t a l s b o t h o p e r a t e smgI1, h i g h -c o s t p l a n t s i n Japan.
Shin-Etsu Handotai w i l l s h u t down i t s p l a n t a s soon a s Hemlock c a n g u a r a n t e e a s u f f i c i e n t s u p p l y o f p o l y c r y s t a l l i n e s i l i c o n . Komatsu p l a n s t o c l o s e i t s s m a l l p l a n t a s soon a s UCC c a n g u a r a n t e e a h i g h -q u a l i t y s u p p l y of p o l y c r y s t a l l i n e s i l i c o n u s i n g t h e s i l a n e p r o c e s s . UCC developed, with the aid of DOE funding, an inexpensive method of converting metallurgical-grade silicon to silane. This process is detailed below: Technical expertise of the Komatsu Electronic Metals Co. has been licensed to decompose the silane thermally to semiconductor-grade polycrystalline silicon in a chemical vapor deposition reactor. A new plant with a capacity of 1200 metric tons per year usirig the Komatsu deposition reactor is expected to start production late in 1984. It is situated at Moses Lake, Washington. Activities for the designing artd planning for an additional 3000-metric-ton capacity plant have also recent1.y been announced; it will be installed in Washington in the late 1980s. Union Carbide is also conducting research on a fluidized-bed reactor jointly funded by DOE. Fluidized-bed reactors will be one of the options to decompose silane for the 3000-metric-ton plant.
The Tokuyama Soda Co. of Japan has licensed expertise from GE (Great Western).
Tokuyama will esseqtially duplicate the GE plant in Japan.
The People's Republic of China has nine operating polycrystalline silicon plants. The largest is at Loyang, Hunan, and has an annual capacity of 25 metric tons. China is dissatisfied with all nine plants and is purchasing equipment to build a new, larger, and more cost-effective Siemens process plant.
Any trade in polycrystalline silicon between the free world and the USSR has essentially been stopped. Although the USSR seems to have sufficient polycrystalline silicon, it does not have the technology to convert it into high-quality integrated-circuit wafers.
The consortium of Allied Chemical Co., Eagle-Picher Industries, Inc., end General Atomic Co. built a laboratory-scale pilot plant to convert by-product SiF4 from the fertilizer industry to silane and to use a fluidized-bed reactor to decompose the silane to semiconductor-grade polycrystalline silicon. This small pilot plant was shut down and disassembled early in 1983. The technical expertise for this process is now for sale.
Three other laboratory-scale pilot plants in the United States use proprietary processes that could potentially manufacture low-cost, high-quality polycrystalline silicon. Two of these plants are based on silane and fluidized-bed decomposition. The third plant uses the hydrogen reduction of SiHBrg. It is expected that at leagt two of these processes will go into production by 1988.

C. MARKET PRICING OUTLOOK
The market price of refined polycrystalline silicon depends on the supply-and-demand scenario, as with any other commodity. Nearly all of the material purchased is used by the semiconductor industry for the fabrication of integrated circuits, transistors, power diodes, and other electronic devices. The demand for photovoltaic devices has been very small, but is expected to increase steadily.
As much as 80% of all polycrystalline silicon is sold under long-term contracts that specify maximum and minimum yearly sales quantities that can or must be delivered or accepted. Such contracts can be broken oniy upon payment of specified penalties, and they specify yearly prices based on agreed-upon escalation scales. Few polycrystalline silicon producers would risk building a new plant without such signed contracts. Year-to-year silicon prices tend to be quite stable once a contract is in effect.
As much as 10% of all polycrystalline silicon is sold on the basis of one-year contracts under which the silicon is purchased in one year for delivery during the next year. During the past four years, silicon prices set on this basis have been far more volatile than those based on long-term contracts.
Less than 10% of all polycrystalline silicon sales take place on the spot market. Producers and long-term contractors take part as sellers in this market. During the past four years, prices on this market have been very volatile. For example, in July 1978, Smiel Co. sold 40 metric tons of prime semiconductor-grade polycrystalline silicon for $40/kg. In July 1980, Great Western sold 30 metric tons of prime semiconductor-grade polycrystalline silicon for $125/kg. In July 1982, Dynamit Nobel and Great Western each sold 30 metric tons of polycrystalline silicon for prices between $38 and $42/kg. Early in 1984 the spot-market prices jumped to about $64/kg for short periods.
There are two types of "off-grade'' polycrystalline silicon that are also sold on the spot market and tend to confuse the situation. This off-grade material, although not suitable for the most exacting uses such as manufacturing thyristors or random-access memories, is eminently suitable for use in the manufacture of solar cells, auto rectifiers, and as substrates for epitaxial wafers. This material accounts for approximately 5% of all virgin polycrystalline silicon manufactured. It generally does not meet the specifications of prime-grade pol.ycrystal1ine silicon because carbon, phosphorus, or boron levels are slightly too high. It is sold, as manufactured, to any interested buyer.
Silicon single-crystal Czochralski growth typically results in only 60% of the p~lycr~stalline silicon feedstock emerging as sellable single-crystal wafers. The remaining silicon is scrapped or recycled. Most of this crystal scrap can be recycled and is suitable for regrowth into single-crystal material, which is then processed into wafers by the large crystal-.growing companies for solar-cell manufacturers. Some of the crystal scrap is also made into wafers suitable for the manufacture of other semiconductor devicas or is sold to solar-cell manufacturers as polycrystalline feedstock. lhe prevailing prices for the sale of this scrap in the past have varied between a quarter and half of the price of prime semiconductor-grade polycrystalline silicon.
Not all polycrystalline silicon reaches the market for sale as bulk polysilicon. Much of it is used by the manufacturers to produce wafers for sale. Table 2-6 gives a history of this wafer manufacturer usage of polycrystalline silicon. Table 2-7 predicts the growth of this internal use.
The semiconductor silicon weighted average selling price in previous years (current-year dollars) and the expected price in the future (1982 dollars) are plotted in Figure 2-1. It is forecast that the weighted average price in the future will decrease, based on these factors: ( 1 ) Adequate polycrystalline silicon supply is expected in the future, with the Siemens process improvements now being incorporated by industry. The capacity of the polycrystalline silicon manufacturers is growing faster than projected market demands.
(2) Union Carbide and other U.S. manufacturing teams will enter the polycrystalline silicon market using new low-cost processes to produce a low-cost, high-quality product. 'Ihese new entries can successfully enter the market only with prices lower than the competition or by offering a better product.
( 3 ) It is becoming evident that silicon-integrated circuits will be the electronic workhorse for the rest of this century. Such technologies as silicon on sapphire, Josephson junction devices, magnetic bubble memories, and gallium arsenide devices will have little effect on the silicon device market. Accordingly, polycrystalli~~e silicon manufacturers are now convinced that the silicon market is here io stay and are willing to make the investments necessary to build the new and larger plants required for the future market.
At present the photovoltaic module manufacturers either buy polycrystalline silicon to grow and slice their own Czochralski wafers or buy silicon wafers directly. Scrapped-out pr'ycrystalline silicon is now used primarily to fabricate solar cells. Tni trend will probably continue in the near future. ( 1 ) C r y s t a l g r i n d i n g , 3%.
( 2 ) Non-single-crys t a l l i n i t y , 2%.
( 3 ) High 0 2 and c a r b o n c o n t e n t , 3%.
( 4 ) C r u c i b l e t a i l i n g s , 7%.
( 5 ) C r y s t a l bottoms and t o p s , 10%.  Approximately 5% of all virgin polycrystalline silicon produced does not meet the specifications required for the manufacture of integrated circuits or power devices. The levels of phosphorus, boron, or carbon are critical for integrated circuits, but often less so for solar cells. This off-grade product is available to the solar industry for a price that is usually less than half the spot-market price for prime-grade polycrystalline silicon.
World photovoltaic shipment history and a forecast are given in Table 2-8 for silicon solar arrays. Japan is aggressively capturing a11 increasing market share, primarily because its industry eLcels in volylile production of amorphous silicon solar cells, now used primarily for consumer products.
Silicon consumption history and a forecast for the manufacture of solar cells through 1988 are given in Table 2-9. Solar cells fabricated from Czochralski crystal and semiconductor-grade polycrystalline silicon uscally have at least a 13% to 14% efficiency. Tnis should increase to 14.5% to 15% by 1983 and 15% to 15.5% by 1985. Solar cells fabricated on polycrystalline ribbon have demonstrated efficiency near 15%. Single-crystal ribbon cells have shown efficiency higher than 16%. It is estimated that ribbons will be used to supply approximately half of the photovoltaic market beyond 1987. As can be seen in Table 2  b~o r e than 4 Mid of the shipments were concentrators that use 1% to 2% of the silicon material needed for an equivalent flat-plate power rating.
CNearly 4 MW of Japan's 1983 shipments were small consumer-product power modules, mostly using amorphous silicon cells.

DEPARTMENT OF ENERGY SILICON REFINING PROCESS RESEARCH
In this section, the research conducted by FSA and the Solar Energy Research Institute (SERI) is reviewed. This research has been funded by DOE to develop the technology for refining processes capable of producing low-cost semiconductor-grade polycrystalline silicon. A probabilistic analysis of the potential for the UCC and Hemlock processes to achieve the National Photovoltaics Program cost goals is summarized, and the impacts of the Program upon the commercial pol.ysilicon industry are discussed.

A. HISTORY OF THE SILICON REFINING RESEARCH TASK
The Low-Cost Solar Array Project (LSA), later renamed Flat-Plate Solar Array Project (FSA), was started by JPL in 1975 as a key effort in the U.S. solar energy research program. lhe goal of the project was to demonstrate the technology for low-cost terrestrial p>f~tovoltaic solar arrays. The project was originally sponsored by the Energy Research and Development Agency (ERDA) and is now funded by DOE. It was structured into several technical areas, each given a cost goa,l allocation to achieve the project goal of $0.50Ipeak watt (1975 $). The Silicon Material Task was assigned the responsibility for demonstrating the technology for low-cost polycrystalline silicon processes. fie criteria were that the polysilicon must be suitable for fabricating photovoltaic solar cells meeting the project's energy conversion performance requirement and that the 1986 market price be <$lO/kg of silicon in 1975 dollars. The task plan had two major objectives: (1) To develop the technology for, and to establish the practicality of, processes capable of meeting the guidelines for production, energy use, and economics. (2) To develop information correlating the effects of impurities on photovoltaic cell performance, which could be used to define the purification requirements and the economics of the processes.
'Ihe plan was divided into phases as follows: (1) Determination of the technical feasibility of candidate processes by evaluating chemistry and chemical engineering data obtained from laboratory-scale experiments.
(2) Evaluations of the potential practicality of the processes by analyses of data obtained in scaled-up reactors and of information from preliminary process design calculations.
( 3 ) Assessments of the performance characteristics of pilot-plant-scale Experimental Process System Development Units (EPSDUS) under simulated steady-state operation. The operational data were to form the basis for the final judgment of the capability of a process to achieve the task goal and concurrently to provide information for the equipment and process designs for large production plants.
The SERl polycrystalline program objectives are to conduct research and studies of potential innovative refining processes for production of low-cost polycrystalline silicon.
The process developments and the primary studies of impurity effects were made at JPL under contracts with industry, universities, and non-profit institutes. Support in areas of chemical engineering, impurity effects, material and cell measurements, and economics was given in similar laboratories. Several consultants were used for analyses, critiques, and criticalproblem reviews. The JPL staff conducted theoretical and experimental studies in JPL laboratories in the areas of material characterization and chemical reactor research. The staff also performed technical evaluations and managed the contracts.
Twelve polysilicon processes were investigated at JPL, most of them to an extent that technical evaluations of their potential to meet the task goal could be made. Devk?.opments were eliminated from the task program by decisions based on comparative technical evaluations when the task plan necessitated reductions in scope. In some cases, the potential economic and energy-use advantages of the eliminated processes were sufficiently encouraging that further development was continued under private funding. Examples are the bromosilane process of J.C. Schumacher Co., sodium reduction of silicon tetrachloride by the AeroChem Research Laboratories process, the direct-arc furnace process of Dow Corning Corp., and sodium reduction of silicon tetrafluoride by the SRI International, Inc. process.
In each of these examples, chemical engineering studies to characterize the main reactor, to establish product purity, and to design a process remained uncompleted when the funding withdrawal decisions were made. The two processes that were the most extensively developed by the task were the Hemlock dichlorosilane (DCS) C M process and the UCC silane process. A hydrochlorination reactor for converting metallurgical-grade silicon feedstock and recycled SiC14 into SiHC13; units for removing metals, and a redistribution reactor using an amine catalyst for converting SiHC13 into SiH2C12 are common to both processes. B.

HEMLOCK SEMICONDUCTOR CORP. DICHLOROSILANE CHEMICAL VAPOR DEPOSITION PROCESS
The research and development of the Hemlock process was carried through the phase of the studies characterizing the operation of a process development unit for converting trichlorosilane into dichlorosilane at a rate of 100 kg/h. The conditions for the deposition in a modified stainless-steel chemical vapor deposition reactor (with cooled walls) were to be established, meeting the contract ~~3 1 s of a rate of 2 g/h per cm rod length and 40% conversion yield. The goal for energy use of 60 kWh/kg was not attained, the average experimental value being 90 k~h/kg. The product was shown to be equivalent to semiconductor-grade silicon. However, the data developed, which describe the flammability and explosive properties of DCS, showed it to be far more haza~rdous than TCS or hydrogen. This concern forced a redesign of the process so that no DCS storage is allowed. The DCS ( S~H~C~~) is diluted with hydrogen directly after distillation and then is fed into the deposition reactors. The decision for the commercialization of this process based on the Hemlock process flow diagram shown in Figure 3-1 (~eference 8) will depend on trade-offs involving the advantages of conversion yield, deposition rate, energy use, and the disadvantage of operating without a DGS stora.ge tank.
The chemistry involved in the DCS CVD process is well established, as is the reactor technology. The stoichiometric equations are given here. The deposition reaction yields are approximately 15% HSiC13, 12% K2SiC12 and 40% Si.

Hydrogenation of Metallurgical-Grade Silicon and Silicon Tetrachloride:
Dichlorosilane Synthesis :

C. UNION CARBIDE CORP. SILANE PROCESS
The DOE-JPL-sponsored research and development of the silane process by UCC was structured in three phases. In Phase I, the engineering feasibility of the process was proven by determining the conditions for silane (~i~4 ) production from mgSi in a small-scale process development unit and by showing that the free-space reactor and the fluidized-bed reactor were both candidates for the silicon deposition step. A complete process flow sheet was developed from engineering process design studies. Then, in Phase 11, a detailed process design package was completed for a 100-metric-ton/year Experimental Process System Development Unit (EPDSU). Phase I11 was for the design, installation, operation, and evaluation of the EPSDU and for the continued development of a continuous deposition reactor for low-cost, large-scale production.
The silane process is a two-stage process. Reaction steps for the conversion of mgSi and recycled silicon tetrachloride (sic14) and H2 into SiHC13 (TCS) and then the sequential redistribution of SiHC13 into SiH2C12 (DCS) and SiH4 are in the first stage. The thermal decomposition *Qualitative equation only; exact reaction is temperature and composition dependent. The t r a c e m e t a l c o n t a m i n a n t s , PCl3, AsC13, NiC12, CuC12, and CaC12, h a v e a h i g h e r S o i l i n g p o i n t t h a n t h e TCS and t h u s w i l l b e r e j e c t e d w i t h t h e STC and e v e n t u a l l y removed a s s l u d g e from t h e s e t t l e r t a n k .
I f a n y of t h e s e m e t a l s a r e p r e s e n t a s h y d r i d e s , t h e y w i l l s t i l l b e w i t h t h e TCS.
Source: F i n a l Report, Low-Cost S o l a r Array P r o j e c t , C o n t r a c t No. 954334, Union Carbide Corp. The TCS, which also has monochlorosilane and DCS mixed with it, is fed into the third column, along with a second feed from the recycle of the silane column. This is set to deliver 97% of the TCS to the bottom and 97% of the DCS to the top as distillate. The bottom product is cooled and fed through a Rohm and Haas Amberlyst A-21 amine-base ion exchange resin in the first catalytic redistribution reactor. This reactor catalyzes the following reactions : The predominant product is DCS, the yield being about: 127,. The product stream from this redistribution reactor is fad t.o the third column to separate the DCS from the STC, which is recycled.
The DCS is fed into a second redistribution reactor containing the same catalyst to produce about 11 mole % SiH4. The product stream is fed into the fourth column, which is designed to remove B2Hg to <0.010 parts per billion and thus to yield very pure SiH4. The SiH4 is stored as a liquid and fed as a gas to the deposition reactors. The TCS mixture from the final column is returned to the third distillation column.

SILANE-TO-SILICON CONVERSION PROCESS
The research for the development of a fluidized-bed reactor (FBR) for silane (siH4) decomposition is being conducted by both JPL ~ild UCC. The primary objectives of the JPL research are to characterize silicon deposition at high SiH4 concentrations, to develop a means for seeding the bed, and to determine the required fluidized-bed operating conditions to produce semiconductor-grade silicon. The UCC program is focused on establishing the conditions for steady-state operation at a SiH4 concentration of about 25% and for ensuring semiconductor-grade product purity. Considerable progress has been made in both programs. At JPL the reactor characteristics have been studied using SiH4 concentrations of 20% through 100%. The operational conditions have been adjusted so that greater than 90% silicon deposition in the bed has been obtained, even at the highest silicon concentrations. Good 'eposition rates occurred; in the run at 80% SiH4, the deposition rate was j.5 kg/h. At UCC, operation was maintained continuously for 44 h using a 20% SiH4 concentration. However, unacceptable contamination of the product has occurred in both reactors due to silicon particle abrasion of the metallic reactor walls. Liners of quartz and silicon will be inserted in the reactors to prevent contamination in future runs. The full. establishment of the silane process as a low-cost polysilicon process meeting the FSA economic goal depends on the completion of the development of the fluidized-bed deposition technology. E.

NEW POLYSILICON REFINING PROCESSES
JPL has a contract with Energy Materials Corp. for the development of an electrochemical silicon refining process that uses mgSi feedstock: solar-grade silicon is deposited on a graphite cathode from a Cu3Si:Si anode by electrolysis in a molten-salt cell. The silicon-copper alloy and the electrolyte are contained in a vitreous carbon crucible at temperatures of 750°C to 8000C. The alloy is formed from m,?Si and 99.999% pure Cu containing 1 ppm Fe as the majority impurity. The electrolyte is a molten mixture of Suprapur* lithium fluoride and Suprapur potassium fluoride with KzSiF6 added as the transport agent. The K2SiF6 is synthesized in situ by reactively dissolving highpurity SiF4 in a KF-rich melt. At operating temperatures, the silicon-copper alloy is a two-phase solid with a primary silicon phase embedded in a CugSi matrix. This facilitates the casting of silicon plates, a prerequisite for an efficient, low-cost electro-refining process. The use of this composite as i... anode enhances the purity of the refined sil.icon by about two orders of magnitude. This enhancement is due to the semipermeable nature of the CugSi phase. This process was originally researched at SERI and has been shown to be an efficient and potentially cost-effective process for producing solargrade silicon.
SERI is investigating the refinement of mgSi by a chemical vapor transport filtration (VTF) process. In this process, a Cu3Si:Si anode reacts with hydrochloric acid (HC~) vapor to produce chlorosilanes which are then decomposed on hot silicon filaments to produce silicon. The silicon transport rate in this process has been measured and the results support the hypothesis that the transport may be limited by diffusion rate of the silicon in the anode. Experiments are under way to determine the activation energy. A large rer , which permits the use of considerably larger anodes, is planned.
A single-crystal silicon boule with (111) orientation was grown from 75 kg of VTF silicon material using a Czochralski crystal-growth technique. Devices made from this crystal measured a total area (0.1 cm2) efficiency of 9.8%. The control cells had an average efficiency of 9.6%, with no cell having Voc greater than 604 mV. These results indicate that the VTF-refined silicon IS of good quality for solar-cell applications.

F. PROBABILISTIC ANALYSIS OF SILICON COST FOR UNION CARBIDE CORP. AND HEMLOCK SEMICONDUCTOR CORP. PROCESSES
A probabilistic analysis (Reference 10) was performed at JPL, using the SIMRAND (Simulation of Research and Development) model and the Improved Price Estimation Guidelines (IPEG) methodology, both developed by JPL. SIMRAND is a Monte Carlo simulation model that can perform algebraic operations on probabilistic inputs. This analysis assessed the probability that the new refining processes can achieve the current DOE cost targets of less than $20/kg with a goal of $16/kg (1982 $1.
-*Trade name of Merck Co., Germany, for fluoride chemicals.
The IPEG methodology was used to provide cost estimates for manufacturing facilities. The IPEG results using identical inputs were compared with the data from the cost analyses performed by Lamar University and the Texas Research Engineering Institute to verify this methodology for a chemical processing plant. On the average, the differences were about 2.6%.
The IPEG equation includes inputs for plant investment, equipment costs, materials costs, utilities costs, labor costs, and output quantity. For this study, a reference output capacity of 1000 metric tons per year was chosen for use throughout the analysis.
This cost-estimating methodology was then incorporated into the probabilistic analysis. Probabilistic distrik-ltions were used as inputs to the IPEG equation. In the case of equipment cust, the processes were separated into sections and distributions were encoded for each one. Each section represented a step in the silicon purification process. Similarly, utilities cost was encoded by type (i.e., electricity, steam, etc.).
These were aggregated to produce the input. For materials costs and labor costs, the total input for the process was encoded directly because breakdown by process step was more difficult. This strategy captured the essence of the economics of the processes while keeping the data encoding and analysis task workable. Items such as recycle loops were not modeled separately, but their influence on plant economics can be seen through their impacts on other inputs.
A distribution for each variable was initially encoded based on data from experts within JPL. In a second encoding round, the dcta were refined and inputs were included from the companies developing the processes. For those cases where the inputs of industry and JPL differed, a compromise distribution was developed. For some variables, proprietary restrictions forced the industry representatives to limit their inputs to brief comments on the JPL distributions. Seventeen of these distributions were encoded using this methodology. These data formed the basis for the remainder of the analysis.
To perform arithmetical operations using probabilistic inputs in a practical manner, the SIMRAND model was used. The Monte Carlo simulation model can include equations and perform the necessary operations using probabilistic inputs.
Some economic assumptions had to be made to perform this analysis. The assumptions were: an equipment lifetime of 10 years; income taxes and property taxes of 50% and 2%, respectively; business investment tax credits, rate of return on equity of 20%; and contingencies of 15% of the total equipment cost. The remaining assumptions are numeroils, but do not have a large effect on the results. price expected for the UCC silane processes (using the FBR and Komatsu reactor as labeled) and the Hemlock process. The curves extend, from lower left to upper right, from the most optimistic to the most pessimistic scenarios for each technology. The vertical axis represents cumulative probability from 0% to 100%. A point on the curve portrays the probability that the technology content and process steps can be used for calculations of the functional sensitivity of cell behavior to these factors. Tnese relationships can, in turn, be used for the determination of the requirements for the polysilicon composition and the appropriate processing procedures for defined cell specifications. Extensive experimental data rere obtained and the effects of impurities, process steps, and impurity process integration were described by equations. Extensive contract effort was made by the Westinghouse Research Center under the direction of R. Hopkins (the contributions of the ].ate R. Davis as a principal investigator in the theoretical analysis and in structuring the studies were especially notable). The data and correlations were used by A. Yamakawa of JPL (Reference 11) to devise a elide rule to identify the effects of impurities on solar-cell efficiency, taking into account various procedures for converting the polysilicon into simple crystalline ingots or ribbons. This experimental information and the theoretical investigations done by Professor C.T. Sah, under JPL contract, relating the effects of impurity content and cell design on cell characteristics, have formed the basis for the research program to develop high-efficiency cells.
Another benefit from the impurities studies is the development of the requirements for a solar-grade polysilicon material with specific impurity levels that would be acceptable to produce a desired solar-cell efficiency.
The specification for the solar-grade polysilicon would be different from the present commercial semiconductor-grade silicon. Based on specific impurity levels of the solar-grade material, solar cells can be produced having a desired efficiency. For the highest possible solar cell efficiency, this solar-grade polysilicon may have impurity requirements that are even more stringent than those specified for the present commercial semiconductor-grade silicon.
Research funded by DOE has been instrumental in developing a number of technically feasible low-cost silicon refining processes that are now used or are expected to be used by the commercial refining idustry. The achievements, which can be attributed directly to National Photovoltaics Program research, have resulted in or w3.11 result in many millions of dollars of cost savings to the commercial refining industry for the production of semiconductor-grade polycrystalline silicon.
In summary, the key DOE-JPL silicon refining research and development achievements are: (1) A process step for preparation of trichlorosilane .using metallurgical-grade silicon, recycled hydrogen, and silicon tetrachloride.

( 2 )
A process for the production of low-cost, very pure silane.

3
An integrated silane process for the production of very pure silicon.
(4) An integrated dichlorosilane chemical vapor deposition process for the production of semiconductor-grade silicon.

(5)
The basic fluidized-bed technology steps for the silane to low-cost silicon deposition.

SUMMARY AND CONCLUSIONS
Over the past five years the JPL Photovoltaics Program PA61 Canter has conducted silicon refining industry surveys to establish the current and future availability of refined silicon and the product market price the photovoltaics industry is and will likely be paying for the fabrication of solar cells.
Another objective of these surveys was to assess the plans of the silicon refining industry and industry reactions to the technology developments funded by DOE to provide low-cost polycrystalline silicon for the photovoltaics industry. Tne surveys were conducted primarily by Remo Pellin, a JPL consultant.
Many industry contacts have been established in these surveys and the industry has been cooperative, friendly, and very active in keeping up with state-of-the-art development activities to improve the quality of the semiconductor-grade silicon produced, to increase yield, and to reduce cost.
Although some of the data presented in this report indicate that plant capacity is in excess of demand, the industry has closely matched production to market demands. There were, however, some periods during which polysilicon for spot-market purchases was in short supply and, as a consequence, spotmarket selling prices skyrocketed.
The years 1980 through most of 1983 were low sales years for polycrystalline silicon. The late-1983 boom in sales caught all producers by surprise and, early in 1984, annual production was not up to demand. In this period, polished-wafer companies were taking steps to increase production substantially. Normally these companies recycle about 10% of the scrap polycrystalline silicon back into their processes. Because this practice reduces polished-wafer yields somewhat, it has been minimized for the present. Eventually all scrap silicon may be used for recycling at the crystal-growth stage. The discontinuance of scrap recycling factor has increased near-term polycrystalline silicon wafering supply requirements by 10%. All polished-wafer users are more optimistic about their business now but are concerned about a possible near-term polycrystalline silicon shortage. On the average, most companies are ordering 10% more polycrystalline silicon than is required for day-to-day needs in order to increase inventory.
The use rate of polycrystalline silicon in e<irl.y 1984 was at least 20% higher than expected. In early 1984, the world pol ycrystalline silicon production rate was about 4000 metric tons per year, and based on a forecast demand of 3960 metric tons for 1984, it can be assumed that sporadic short-term shortages may develop. Because Wacker Chemetronic GmbH and Hemlock Semiconductor Corp. will increase capacity by process changes and Union Carbide Corp. will come on stream late in 1984 with a new plant, polishedwafer manufacturers will come to accept, temporarily, shorter inventories and will use some scrapped-out polycrystalline silicon. Thus enough polycrystalline silicon should be available for all, although some transient conditions of limited spot-market supply may persist.
Today and in the near future, most polycrystalline silicon will be manufactured by Siemens-and modified-Siemens-process plants. Wacker and Osaka Titanium Mfg. Co. will use trichlorosilane as the silicon-so~lrce ingredient. Hemlock uses trichlorosilane now and may eventually use dichlorosilane. UCC will use silane for production of silicon.
Research and development projects sponsored by DOE and managed by JPL have developed improved processes and procedures for producing lower-cost polycrystalline silicon manufactured by variants of the Siemens process. Research sponsored by DOE/JPL and performed by UCC and Solarelectronics, Inc. on the hydrogenation of silicon tetrachloride to trichlorosilane has now been applied to full production at Wacker and soon will be at Hemlock and Osaka Titanium. n.is research showed how by-product silicon tetrachloride could be reduced inexpensively to trichlorosilane in the presence of metallurgicalgrade silicon. Tne process is shown as Reaction 2 in Figure 2-1. rile process modification at Wacker apparently reduced the manufacturing cost of polycrystalline silicon from over $30/kg to about $24/kg. It is expected that this process step will be used by all polycrystalline silicon manufacturers for the rest of this century, regardless of the exact process definition.
The Hemlock polycrystalline silicon plant might have been forced to shut down if JPL-sponsored research had not shown a way to continue potential profitability, by modernizing with new process steps such as hydrogenation of silicon tetrachloride to trichlorosilane. With this change, Hemlock should be able to produce polycrystalline silicon at a cost competitive with Wacket's, despite disadvantageous electric utility and labor costs.
Perhaps the most classic example of Government-sponsored high-risk research, showing a promising result and involving industry commercialization of the process, is the UCC process. UCC had considered entering the polysilicon business several times before 1980, but did not make the commitment until the JPL program showed technical feasibility. With a 120-metric-ton polysilicon plant in operation, a 1200-metric-ton plant scheduled to start up late in 1984, and a 3000-metric-ton plant announced for 1988, UCC will help make the United States a Jominant force in world' production of refined silicon. The new UCC Moses Lake, Washington, plant will require an investment of $100/kg of annual capacity. The great majority of this investment is in Komatsu reactors for the final decomposition step. Thus, even plants with an annual capacity of 1200 or 3000 metric tons are expensive. The fluidized-bed deposition reactor, if successful, should greatly reduce the capital cost when installed in future silicon refining plants.
DOE-funded research has been instrumental in developing a number of technical achievements and integrated approaches for low-cost refining processes that are or are expected to be used by the commercial refining industry. Achievements that can be directly attributed to the PV program research have or will result in many millions of dollars of cost savings to the commercial refining industry for the production of semiconductor-grade polycrystalline silicon.
It can be said that the overall DOE-sponsored silicon materials research program has had a major effect on both the world's polysilicon producers and silicon supply. It is one area in which the United States, through proper use of Government funds, has maintained technology leadership.