Materials issues in 2000/2005
11/01/1998
Materials issues in 2000/2005
Ed Korczynski, Senior Technical Editor
The demand for semiconductor devices is continuing to increase at such a rate that the number of devices produced worldwide in the year 2005 is expected to be 10? 1996 levels. The amount of materials consumed, and the waste materials discharged, could increase proportionally. Resource conservation and other environmental issues, however, mandate that the expected increase in output must be accomplished without a proportional increase in material consumption or waste generation.
Since a supply shortage in a single component sector - such as epoxy or polysilicon - can slow the entire semiconductor industry, the Materials Savings Subcommittee of SEMI Japan began an investigation by identifying potential shortages from a global perspective. Fortunately, it appears that there are no significant supply-side limits, even if chip production increases by 10?, partly because chip production and the amount of the material consumed are not necessarily in a one-to-one relationship. Their findings are summarized below.
Gases
The supply of raw materials or gases for the manufacture of semiconductors is dependent on other industries that consume large quantities of the same resources. Since these other industries will also continue to require raw materials, there will probably be no direct raw material or gas supply problems. Supply problems could arise, however, if the material-producing industries grow slowly or shrink.
For example, the steel industry provides the argon gas that is widely used in single-crystal silicon growth (see "Silicon wafers" section on p. S8). Since the growth rate for steel is virtually zero, a large increase in argon consumption could lead to a supply shortfall. Consequently, the expensive argon must be recycled during production to reduce overall manufacturing costs.
Chemicals
There will be increasingly tighter chemical purity requirements, necessitating early determination of specifications and development of analytical methods. Otherwise, a supply crunch could occur in ultrapure chemicals (Fig. 1). Because pollution regulations will become increasingly stringent, recycling and reduction of chemical consumption are critically important. Re-using chemicals, however, is a difficult task. Some volatile organic solvents, such as isopropyl alcohol, are now trapped, purified, and recycled. For inorganic chemicals, the only practical approach may be to reduce the amount used.
Among the chemicals that could present resource problems is hydrofluoric acid (HF). The production of fluorine in Japan ceased long ago, and all current supplies come from imports. Fluorine could be recycled through the use of calcite (CaCO3) to recover CaF2.
 |
Figure 1. Chemical delivery system changes from the 1980s to the early 21st century.
 |
Deionized water
Table 1 shows that device shrinks have been accompanied by a quantum leap in the quality of deionized water (DIW). As the volume of DIW consumed/fab has risen by 2?, recycling has become an effective way to minimize consumption. Currently more than 90% of all fabs are recycling more than 70% of their DIW. Barring a radical change in technology, the amount of DIW used/fab in the year 2005 will reach 250 m3/hr.
Silicon wafers
If 25 gm of single-crystal silicon is used for every 100-mm wafer, the total single-crystal silicon requirement in the year 2005 will be 5970 tons/month. If the yield from the single-crystal production process is 60%, the amount of single-crystal required will be 10,000 tons/month. Allowing for maintenance and other overhead, and for a 10% surplus inventory level, the required monthly single-crystal tonnage will be 11,000. Although 5% of the current production level comes from float zone (Fz) silicon, this calculation assumes that all single crystals are produced using the Czochralski (Cz) technique. Based on the assumption that the current trends in wafer production will continue unchanged, the report predicts no significant supply problems in wafer production.
The projected 10-fold increase in wafer output may, however, present problems in the supply of electricity, metallurgical silicon, and argon gas. Among these, the supply and cost of electricity seem to present the greatest challenge (Table 2). The electricity may come from nuclear energy, thermal power plants, or hydroelectric stations. Brazil, Norway, and China are good locations for power plant construction due to raw material excavation, transport, electricity cost, high-quality coal supply, and labor cost considerations.
Most metallurgical silicon is used in the steel and aluminum industries, which account for two-thirds of the demand. The
remaining one-third of the output is used in the chemical industry. Output during 1995 was 780,000 tons, of which approximately 48,000 tons were used for the production of trichlorosilane. Raw material that is low in phosphorus and arsenic content is used for the production of metallurgical silicon.
Since no new electric furnaces were constructed in recent years, there will presumably be no increase in global metallurgical silicon output from 1996 levels. If IC production trends continue, the 480,000-ton demand for silicon wafers will no longer represent a "special case" demand. It may no longer be possible to treat purity of the output as a priority goal, necessitating techniques to remove phosphorus and arsenic impurities.
To meet a 10-fold increase in the output of wafers, 108 ? 106 kg/month of argon will be used for single-crystal growth, which is 56% of the current worldwide argon gas production capacity of 192 ? 106 kg/month. For the production of argon, the dependency of the semiconductor industry on the steel-making industry may deepen. Serious efforts must be made to increase the rate of argon recycling.
When the output of silicon wafers increases by 10?, the semiconductor industry will have moved from a position as a
secondary user of raw materials to the position of a primary user. The selection of new wafer process construction sites, improved energy conservation in the processes, and improvements in equipment, technology, and yield will all become necessary.
A rapid increase in demand and a sudden elevation in specification levels could lead to either temporary or persistent supply shortfalls. We need to move away from the conventional wisdom that a 10-fold increase in output requires a 10-fold increase in input.
Quartz, silica glass, and related materials
Quartz glass is used for silicon single-crystal-growth crucibles, heat treatment tubes, high-temperature boats, jigs, and masks (Table 3). Although the industry has experimented with several other heat-tolerant materials, they have not taken hold because of purity and production problems, including the difficulty of making adequately large crucibles, boats, and other structures (Fig. 2).
Reaction tubes, boats, crucibles, and cleaning baths are made by flame treatment or high-temperature molding of either natural quartz nuggets or a molten ingot. The original quartz nuggets contain different impurities depending on the place where the material was formed. Silicon is an abundant element of the earth`s crust (second only to oxygen) and is found in combination with other common elements such as aluminum, iron, calcium, sodium, potassium, and magnesium. The specific compositions in which silicon, as quartz, combines with these elements are highly dependent on geography.
 |
Figure 2. Rise and fall of heat-resistant materials used in semiconductor device production.
High-purity raw quartz has become increasingly scarce. To protect the natural environment, including scenic sites, areas that contain high-grade raw materials are not always available for mining. Imports of raw quartz into Japan have increased over the years. Formerly, Brazil was a major supplier of quartz sand for Japan. Now the US, India, and China are leading sources. Unimin Corp. (US) supplies quartz sand from a deposit in the Appalachian Mountains, which can be sustained from 2-30 years at the present mining rate.
If a majority of wafers produced in the year 2005 are 300 mm, the use of 800- to 900-mm crucibles will be essential in order to keep wafer costs low. Synthetic silica glass will also be necessary to meet purity requirements (Fig. 3). Since single-crystal growth requires high temperatures and long growth periods, and the crucible must support >250 kg of molten silicon, one cost reduction solution is to line the inside of a natural quartz crucible with synthetic silica glass. The option of creating a composite crucible by blending quartz glass with highly pure crystalline silica must also be considered.
 |
Photomasks
Synthetic silica glass is a standard material for photomasks because of the necessity to transmit increasingly shorter wavelengths. A move from DRAM to embedded memory production may lead to an increase in the quantity of photomasks used. In addition, there is an increase in the average mass, which translates into a substantial increase in the amount of silica glass.
Due to further increases in chip size and complexity, the use of 9-in. (9 ? 9 ? 0.35 in., or 9035) photomasks will begin by the year 2005. A 9-in. mask weighs approximately 13? more than a 5-in. mask. Therefore, the demand for silica glass for photomasks in the year 2005 is likely to surpass a straight line projection from the current trend. Demand may be 60? the level predicted by Rose Associates.
Structural ceramics
The semiconductor industry has tried, discarded, and then reappraised several candidate heat-resistant ceramics (Fig. 2). Most potential alternatives have failed due to purity, processing, and large-scale construction requirements. Mullite (highly pure aluminum silicate) heat-resistant reaction tubes have been rejected because of purity deficiencies. It is difficult to produce boats and reaction tubes out of silicon because of the difficulty of fashioning the material into large structures. However, silicon`s high purity motivates some companies to use bonding techniques to make boats and reaction tubes with silicon.
Silicon carbide (SiC) is used as a component material through purification, bonding, machining, and CVD coating. Unlike quartz glass, however, silicon carbide cannot be flame-processed. Recently, improvements in the starting powder and in certain processes allow for the casting of large structures. Because of the requirements for high purity and the prevention of heavy metal diffusion, Si3N4 and SiC fabrication technology should be developed by the year 2005.
It is possible to exploit by-products and waste materials generated from silicon production processes as raw materials for the production of silicon carbide and related materials. The challenge remains to build sufficiently large equipment to
fabricate large boats and reaction tubes from silicon carbide and other materials.
 |
Figure 3. Demand for quartz sand for crucible production use in 300-mm wafer growth (fully synthetic or composite natural/synthetic cases).
Lenses and optical materials
The shipment volume of exposure equipment in the last 10 years has fluctuated by ?400 units, with an average of 800 units/yr. In the future, the curve will probably fluctuate around a base of 1000 units/yr. Even if the output of semiconductor devices should increase by a factor of 10, productivity improvements including the use of larger wafers will keep the number of required exposure machines relatively low. Even if 10? more exposure machines were required, there should be no supply problems, provided that requisitions for equipment are issued in an orderly fashion.
Rather, problems arise from the increasing demand for higher lens quality to process larger wafers with shrinking design rules. Optical steppers are now approaching wavelength limits. Post-optical lithography may be required to fabricate <0.1-?m structures, though such solutions remain incomplete and incremental improvements in optical exposure equipment are critical.
Due to increasingly stringent aberration-control requirements, lens designers are demanding ever higher lens material quality (Table 4). Also, wavelength reduction imposes severe limits on the types of lens materials that can be used. The production of large-diameter lenses in a highly reproducible manner mandates that the materials chosen are relatively easy to process. Furthermore, durability is a critical factor to minimize lens deterioration.
 |
Target materials
In a physical vapor deposition process, the term "target material" refers to the metal that is struck by ionized argon gas to produce the metal vapor stream that coats the wafer surface. Among the commonly used target materials are aluminum, tungsten, molybdenum, titanium, tungsten-titanium alloy, copper, and chromium.
Handling is the most significant issue that must be addressed in the field of target materials. A target used for the processing of a 300-mm wafer will measure 500 mm in diameter and weigh 10-20 kg (Tables 5 and 6), and will thus be difficult to handle during target replacement.
Acknowledgment
This article is based on a translation of the Japanese Report of the SEMI PCS-IS Forum Survey Activity. The PCS Forum was led by Chairman Bujiri Kobayashi, and Vice Chairmen Keiichi Shimakura, Yoshitaka Kawasaki, Issei Imahashi, and Richard Dyck. PCS Vice Chairman Yoshitaka Kawasaki (Shin-Etsu Handotai Co. Ltd.) led the Materials Savings Subcommittee.
Shin-ichiro Takasu, the Materials Section editor from SEMI Japan compiled this report, with contributions from Y. Kato, A. Yamashita, T. Ochiai, H. Yaegashi, S. Matsunaga, T. Nakagawa, O. Ohta, H. Tomioka, S. Shimai, K. Kenmochi, N. Takamatsu, S. Isobe, H. Mihara, T. Shokai, and S. Igata. The report, "The Shape of Semiconductor Plants in the Years 2000/2005," was originally edited by SEMI Japan. The English report, translated and edited by O`Mara & Associates, Palo Alto, CA, is published and available through SEMI Technical Publications, 805 E. Middlefield Rd., Mountain View, CA 94043, ph 650/940-7903.
ED KORCZYNSKI is a Senior Technical Editor at Solid State Technology. He received his BS degree in Materials Science and Engineering from the Massachusetts Institute of Technology. He has more than 10 years of engineering and management experience in process development and equipment marketing. His current interests are thin films, process integration, and plasma and vacuum technology. He is a member of the
Materials Research Society. Solid State Technology, 1700 S. Winchester Blvd., Suite 210, Campbell, CA ph 408/370-4833, e-mail [email protected].