The Move to the next Silicon Wafer Size

Transcription

1 White Paper The Move to the next Silicon Wafer Size

2 The Move to the next Silicon Wafer Size: A White Paper from the European Equipment and Materials 450mm Initiative (EEMI450) Introduction: Industry Dynamics and Moore s Law The semiconductor manufacturing industry has undergone dramatic growth during the second half of the last century, and throughout the first decade of the new millennium. A primary driver of that growth is the realisation of Moore s Law, whereby the number of transistors in an integrated circuit doubles approximately every 2 years with an associated increase in circuit functionality, reduction in operational power, and a reduction in unit cost. This, in turn, leads to an increased market demand for consumer products containing silicon integrated circuits, such as personal computers, mobile phones, and other electronic devices. As an economic consequence, the silicon wafers used in the manufacturing processes to produce the silicon chips have undergone a diameter increase approximately every 10 years to improve throughput and reduce manufacturing costs. Silicon wafers that have been used in high volume manufacturing processes range from 1-inch diameter, to the current state of the art diameter of 300mm (11.8- inch, usually referred to as 12-inch). During this wafer size evolution, the ever increasing complexity of the supply chain of semiconductors has, and continues to demand, an increasing involvement from materials suppliers and equipment manufacturers. Tier 1 Initiatives The next silicon wafer size is destined to be 450mm (18-inch), and activities are currently underway amongst the Tier 1 Semiconductor Manufacturing Companies, Intel, Samsung, TSMC and others, to prepare for manufacturing using this next wafer size in the second half of the current decade. In order for this to happen, a plethora of technological breakthroughs are required from the Equipment and Materials organisations, beyond a simple scale-up and extension of current technologies. This represents a global challenge as these organisations operate in all regions of the World. The European Equipment and Materials organisations therefore have a key role to play in this activity. By taking this initiative to be involved

3 in this next wafer size transition, they will have the opportunity to establish share of the 450mm equipment and materials market. Why the transition to the next generation wafer size? There is one fundamental and compelling question to be answered at each of the semiconductor manufacturing industry s wafer transitions; from the very early 2-inch diameter and less generations of the 1950 s and 60 s, through the 3-inch and 4-inch of the 1970 s, the 5-inch and 6-inch of the 80 s, the 200mm (8-inch) of the 90 s, the 300mm (12-inch) of the 2000 s, and now the 450mm (18-inch) generation expected in the second half of this decade and second century of semiconductor manufacturing. That question is: Why is a 50%-diameter increase in semiconductor processing platform needed about every decade? 2inch, 4inch, 6inch and 8inch silicon wafers used in semiconductor manufacturing processes As with the previous wafer size transitions, the move to 450mm is largely driven by the productivity, environmental, and economic challenges of the semiconductor industry as it continues to evolve. In particular: - The number of semiconductor manufacturing fabrication facility ( fab ) construction projects must be sustainable based on both its manufacturing complexity and environmental impact; - The productivity and thus the economic feasibility of semiconductor manufacturing costs must be preserved [normalized for analysis purposes as cost per square centimeter (cost/cm2)] in the face of continuous and exponentially increasing manufacturing facilities, equipment and materials costs. These costs are driven by the technology investments required to meet customer expectations for the doubling of functionality and performance approximately every 2 years, as defined by Moore s Law. If we look at the data governing the volume of silicon shipped during the history of the semiconductor industry (based upon the SEMI silicon shipment history), we can see that, historically, the baseline demand compound annual growth rate (CAGR) is a constant 7.6% since 1993, and prior to that was 15%, with periodic cycles around

4 these core rates. These major business cycles have occurred with a 7 to 8-year periodicity, with a minor slowing cycle in between the major cycle trough and peak. Another factor affecting the growth rate of the industry is the overall downward trend of the Average Selling Prices (ASPs), which in effect has turned what used to be luxury electronic devices into commodity goods for the general public. The chart below, provided by the International Semiconductor Manufacturing Initiative (ISMI), shows what to expect if the CAGR trends continue into the future. At a CAGR of 7.6%, the amount of silicon shipped for all types of electronic products such as memory, microprocessors, and other specific logic devices would double in less than 10 years. This would therefore require new factory capacity to sustain such growth. In previous wafer size generations, this has led to the adoption of the next-generation wafer size in order to reduce the number of factories to be built and sustained. Therefore, at some point in the future along this demand curve, despite any future possible industry consolidation, it will become more economical for chip makers to build one 450mm factory rather than two 300mm factories. At a CAGR of 7.6%, the amount of silicon shipped for all types of electronic products such as memory, microprocessors and other specific logic devices would double in just less than 10 years. When the cycle-based demand growth is applied to the ISMI Industry Economic Model in a scenario without the 450mm wafer generation, the number of 300mm equivalent 35K wafer starts per month (wspm) fab capacity increments rises to above 500 fabs, for all wafer generations, by the middle of this decade, and to above 600 fabs into the next decade. Based on this model, the level of fab capacity in any one wafer generation would rise to a level unprecedented in history, and create untenable challenges for companies required to meet the anticipated customer demand associated with many different products. With further potential industry consolidation, and therefore new factory builds resting with fewer companies, these challenges become even more critical. Even if the size of an individual company s actual site (versus normalized 35K wspm equivalent increments) grows to take advantage of scale, the pressures upon the company resources for personnel, training, and facility infrastructure will be

5 challenging even under the 300mm-only scenario. In addition, the environmental impact of water usage, sewage effluent, and efficient materials and energy usage will be far greater on a per square centimeter basis, as demonstrated by the 300mm wafer generation s own history during the 200mm to 300mm transition from Another consideration in the absence of 450mm manufacturing would be the number of incremental new 300mm fab additions required to support the growing semiconductor market. This would be greater than the number of 450mm facility builds required, and hence could increase the number of annual build projects above the historic steady and sustainable level. A 450mm Roadmap Looking from a cost perspective, historically, the cumulative cost per transistor benefit, weighted across all product types (with memory functionality most heavily weighted) resulted in a compound reduction of cost per function of -29% per year. This reduction was a combination of doubling the functionality in a square centimeter every two years (as per Moore s Law ) and keeping the cost of manufacturing for that square centimeter approximately flat. The benefit of a new wafer size generation s productivity has been estimated in the past by resetting by a 30% cost reduction every 10 years the ~3-4% /year exponentially increasing costs due to technology cycle upgrades and insertions, e.g., copper interconnects replacing aluminium. Although not necessarily desired, the ITRS has been anticipating a slowdown in the rate of the technology cycle from 2 years, to 3 years, resulting in a slowing of the functional density in a square centimeter. In the absence of a 450mm productivity solution, the net slowdown effect upon the combination of the slower density and the higher cost per square centimeter results in only a -27% average cost/transistor reduction rate. This may appear to be a minor effect, but over the timeframe, the slower rate produces a cumulative ~1 trillion dollar productivity difference impact without the 450mm wafer generation productivity gain. (In order to arrive at this statement, an in depth analysis should be performed per product type, taking in to account certain industry specific details e.g. EUV introduction, initial high cost of 450mm silicon and Flash cost growth due to 3D layer implementation. This analysis is possible with the IC Knowledge Strategic Cost Model tool; see below.)

6 Flash ~25% Cost Savings DRAM, MPU ~30% Cost Savings l for EEMI450 white paper ces] [graphic proposal for EEMI450 white paper without $ references] Flash ~25% Cost Savings DRAM, MPU ~30% Cost Savings Source: Strategic Cost Model revision February 2012 All rights reserved IC Knowledge ( for use by licensed subscription owners... over the timeframe, the slower rate produces a cumulative ~1 trillion dollar productivity difference impact without the 450mm wafer generation productivity gain. A New Analytical Model In 2010, an Industry analysis Company, IC Knowledge (ICK), ( developed a new commercial Strategic Model, based on the 2009 ITRS ( which enables cost analysis over the range. A demonstration of the new Strategic Model analysis for the 300mm to 450mm analysis under various assumptions (original 200mm to 300mm assumptions, ISMI assumptions, and ISMI/SEMI assumptions) can be viewed online at the ICK website. The analysis demonstrates the new high levels of cost per cm2 that are a challenge for each product group (memory, logic etc.), and also demonstrates cost per cm2

7 reduction possible from the 450mm generation production at a given technology level (16nm Logic and NAND, and 24nm DRAM are analysed). One generic economic conclusion of concern is that increasing cost per cm2, and reduced density over time will affect the elasticity of demand for semiconductor products and slow down industry growth. Increased costs due to the absence of the 450mm will contribute to the economic slowing, in addition to the fab sustainability and environmental impact. (It is important to note that the analysis snapshot assumes that silicon starting material is compared at a mature cost level. For every wafer generation the starting material starts out very high, but drops quickly as volumes ramp.) Conversion to 450mm diameter wafers can only occur if the economic advantages can be demonstrated for the entire supply chain. One of the benefits of cost analysis using the ICK Strategic Model (available to purchase on the ICK website) is that it generates the demand for equipment to manufacture an incremental fab volume at a particular technology level of a product process flow and assumptions. There are many different sets of assumptions that can be made when developing a model of this type, but the fact remains that despite all of these differing scenarios, as with historic wafer size transitions, the move to 450mm enables a reduction in cost per cm2. Expectations for the industry s future anticipate a healthy demand for both 300mm and 450mm equipment, which should justify the research and development investments by manufacturers, suppliers, and their consortia, as well as funding agencies to prepare and meet the 2015 target ramp (2009 ITRS scenario), and the next years of affordable customer demand. The ability to extend Moore s Law into the future is dependent upon the development and production of new semiconductor manufacturing equipment. The operating costs associated with the silicon IC manufacturing Industry are heavily dominated by the cost of the equipment depreciation and maintenance. Therefore, by providing a new generation of processing equipment for 450mm silicon wafers, IDM, Foundry and OEM organisations are set to benefit in the future, and enable the sustained growth of the Industry. This fact remains true despite future potential industry consolidation and increased use of Foundries by Fab-less and Fab-Lite companies.

8 The operating costs associated with the silicon IC manufacturing Industry are heavily dominated by the cost of the equipment depreciation and maintenance. The equipment of the 450mm generation will have to meet currently unknown high yield manufacturing ramp rates, challenging levels of defect free operation, and extreme reliability performance. Therefore, a series of research and development efforts will have to be invested into equipment improvements and innovations. Of special importance will be the areas of - mechanical handling improvements, - contamination and defect free operation, - joint assessment for fast ramp-up - Advanced Process Control (APC), - comprehensive introduction of virtual metrology, - innovative maintenance strategies, - ultra low relative energy consumption, - severe reduction of consumables and operating materials, their re-use and regeneration (Cost of Ownership improvements), - waste free or waste reduced processing. The above list is certainly not exhaustive, and it is definitely expected to transfer the resulting equipment improvements into the next generation of 300mm production equipment. There exists an outstanding benefit for European Equipment and Materials manufacturers to pool their efforts, to get support by Public Authorities at European and National level on their first steps into this required research and development. Thus, not only the 450mm technology will become a success story, but also the next generation of 300mm equipment and the More-Than-Moore fabs will have strong benefits, resulting presumably in safeguarded employment and in several thousands of new, high quality jobs across Europe. Perhaps to reduce risk, innovations on 300mm equipments and then scale up to 450mm may be a sometimes preferred route for both OEMs and IDMs, and subsequent process technology development on 300mm wafers, then scale up, a cheaper approach for IDMs to follow. The presence of an appropriate supply chain, and of customers of 450mm processed wafers, could be an invaluable attraction to maintain

9 semiconductor production in Europe and to consider Europe for additional 450mm fabs. Background on Equipment and Materials Overview The following chart shows the worldwide sales of semiconductors and semiequipment, plus the percentage of annual turnover the semiconductor manufacturers spend on equipment. The following can be deducted from these figures. Source: Future Horizons/Internet From 1995 to 2010, the worldwide semiconductor turnover increased over time, despite occasional ups and downs, around a factor of 2. This is not the case for their capital equipment investments. While the peak in the year 2000 has still not been reached, the period 2002 to 2007 shows a steady increase, mainly based upon the investments in new 300mm fabs. This becomes clearer when using an appropriate scale.

10 Source: Future Horizons Additionally, one can observe that during this phase the investment in new equipment as percentage of semi turnover was about constant, around 15% (these figures differ somewhat depending on the source, some claiming higher values in the range of 20-22%) after the large millennium peak of It can also be seen that the decrease in equipment sales commenced in 2007, at least a year before semiconductor sales hit the full crisis. Nevertheless, the generally rising investment in equipment in the years 2001 to 2007 can be linked to the number of new 300mm fabs installed during that timeframe. Two data points emerge: Source: IC Knowledge In 2002, 38.2% of equipment purchased by the semiconductor manufacturers was for 300mm production, and 1.4% of silicon wafers processed in 2002 was 300mm.

11 In 2008, 92.1% of equipment purchased was for 300mm production, and 37.4% of silicon wafers processed was 300mm. Therefore it can be concluded: After introduction of the new wafer size, equipment purchases quickly became 300mm dominant. Equipment makers not able to provide 300mm tools could only serve 8% of the total market in Similar scenarios are currently anticipated for the 300mm to 450mm wafer size transition, but further detailed assumption analysis should be performed by the industry working together to provide a more robust and reliable scenario. The Semiconductor Equipment market The semiconductor equipment market is dominated by a few major semiconductor manufacturing companies. The top 5 companies have at least a 40% share of the total worldwide capital expenditures, which in recent years approaches 50%. The top 3 players alone come close to 40% since These 3 tier 1 companies are Samsung, Intel, and TSMC, the same companies that have currently indicated their intentions to move to 450mm production. Source: Company reports/internet It might be expected, and as we have observed for the mm transition, when the 450mm transition arrives at the HVM stage, most of the total investments of these companies will be targeted at 450mm. For equipment makers not participating in the 450mm transition, that will mean, that at least 40% of the total equipment market will not be addressed. That is assuming no other companies will follow the

12 road to 450mm, which will not be the case, as recent announcements already indicate. The mm transition has shown that all companies providing high volume products, such as logic and memory, have moved to 300mm, including the major foundries. With an 8-year delay, we now can see a shift to 300mm by the analog/power providers, such as TI and Infineon. Actually this is a repetition of the mm transition, where slow followers like Bosch are finally moving to 200mm after an 18- year delay. Therefore it can be expected that at least the major foundries, logic, and memory makers have to make a shift to 450mm production to stay competitive, which will result in a total 450mm equipment purchasing power of at least 60 to 70% of the total available market (TAM), and with high invest volumes in the 450mm ramp phase, as we have seen in the 300mm transition. In addition it can be noted that, in general, semiconductor companies have a conservative approach to equipment selection. Installed equipment from a certain vendor will be purchased again during expansions, when performance has been adequately demonstrated. This is based upon considerations such as ease of process transfer, confidence in performance, maintenance know-how, and reduction of spare part inventories. Some companies have even devised rigid procedures in this respect, such as Intel s copy exactly philosophy. For newcomers in the market, a wafer transition is a unique opportunity to enter the market and gain new customers. It is therefore extremely important to be at the forefront of developments to enable introduction of equipment to new customers at an early stage in order to profit from a possible change in vendor policy from 450mm customers, with an attractive offering. The European Equipment Market In this century, the large European based semiconductor manufacturers in total have seen a significant drop in their market position. As a result, the European semiconductor equipment market has decreased considerably in relation to worldwide market, as shown in the next two charts:

13 Source: Future Horizons And specifically Europe: Source: Future Horizons European semi equipment purchases have decreased from 16% of worldwide Capex in 1993, to around 6% in However, sales of European equipment have increased on a worldwide basis over the last years.

14 % WW Sales of EU Equipment Source: VLSI While this is mainly due to the success of ASML, in general one can say that European equipment makers are increasing their market share. These two above-mentioned factors are a clear indication, that EU equipment manufacturers are dependent on, and should target the worldwide market, having a total market share of 20% in 2010, and not so much on the local (European) market with a volume of only 6% of the TAM. According to SEMI Europe, the semiconductor related equipment and materials companies in Europe are directly employing 105,000 employees, which is about equivalent to direct employment by the European semiconductor manufacturers (110,000) and is therefore an important high-tech industry on its own. Materials Looking at the development of silicon wafer production, the following can be observed: the peaks in number of produced successive wafer sizes are spaced 5 years apart until 100mm. 125mm was never a very successful wafer size, and was readily overtaken by 150mm. 200mm followed after 7 years, and 300mm after 10 years.

15 The picture becomes clearer if the produced wafer surface area is normalized per wafer size. In this figure, it can be observed, that both 200mm and 300mm reached the production level of the previous wafer size about 7 years after introduction, although 300mm was at a much higher absolute surface area level, due to increased overall semiconductor production. The introduction of 450mm will likely be in 2015/2016 following the trend that the time increases between introductions of the next wafer size. If the 450mm introduction follows the historical trends, it should reach 300mm production levels in produced silicon surface area in 2023/2024. In the figure below, the pricing per wafer produced unit of surface area is depicted per wafer size over time, with a prognosis for 450mm. According to this analysis, initially every new wafer size starts with high costs, but when mass production sets in, the cost per square unit all converge to the same level, even with an observed increase in wafer quality over time for the larger wafer sizes. During the 300mm

16 transition, the pricing reduction curve is even steeper than in previous introductions. The expectation is that this will also be the case for 450mm, resulting in the conclusion that for the semiconductor manufacturer s substrate costs per unit area will not be inhibitive to introduce 450mm. Source: IC Knowledge In Europe, four wafer suppliers are currently active: Siltronic, Soitec, Okmetic, and Siltronix. The latter two are not major players, and do not manufacture the larger wafer sizes. Siltronic, however, is the number 3 wafer maker with a global presence and a worldwide market share of around 14% in For Siltronic it is important to participate in the 450mm transition to maintain or improve its market share, especially at important customers like Intel, one of the drivers of the 450mm transition. Soitec is producing a special kind of wafer called Silicon-on-Insulator (SOI), which are employed by some logic manufacturers and foundries as starting material to obtain their required device characteristics. During the last years, the market share of SOI wafers of total wafer supply was around 3%. The production in 300mm (in equivalent surface area) was over 50% of total shipped SOI volume. Soitec is the leading supplier for SOI wafers with a market share of 60% in recent years. In order to maintain this position, Soitec is keen to develop 450mm SOI wafers at the earliest possible stage. 450mm: A global effort and co-operation In its history, the semiconductor industry initially went through a number of wafer size transitions lead by individual manufacturers. Because transitioning to the 300mm wafer size was viewed by all to be very costly to be carried out alone, this transition was facilitated through R&D consortia for the first time. SEMATECH established in

17 1995 a global initiative on 300mm transition called I300I. Leading IC manufacturers and equipment and materials suppliers from Europe, the US, and other regions joined this initiative. Japan also established an industry wide initiative called J300. To ensure a cost-effective and timely transition, I300I and J300 collaborated efficiently through 1996 and As a result, the two consortia completed in July 1997 global standards requirements and global joint guidelines on equipment configuration, operation, facilities, automation, FOUP, and carriers. International standards for 450mm carriers, load-ports and developmental test wafers were completed after extensive prototyping and interoperability / cycle testing in cooperative development between component suppliers, SEMI and ISMI. Extensive modeling of 450mm industry economics including tool market projections has been undertaken by ISMI in cooperation with IC Knowledge. The SEMATECH 450mm consortium program has moved to the Albany, NY campus of the College of Nanoscale Science and Engineering (CNSE), and is in the early stages of expanding the test wafer operations to support broad supplier equipment development. Cooperation with the European 450mm Initiative EEMI450 on such development with European suppliers is expected to build on past consultation between the consortia. The entire 450mm tool set must be demonstrated by consortia to support IC manufacturers decisions on launching their pilot operations toward a manufacturing ramp in the next very few years. Shared communication between consortia and companies is hoped to enable the effectiveness of synergistic pre-competitive cooperation. Therefore, a globally-coordinated effort will be needed to ensure a costeffective and timely transition to the next wafer size. 450mm: Imperative for Sustainable Growth In conclusion, based on analysis work by ISMI it is proposed that the 7.6% scenario industry silicon demand growth projection requires the support of a 450mm wafer generation beginning ramp in the second half of this decade. The need for 450mm manufacturing is driven by manufacturing economic, ecological, productivity and sustainability feasibility issues. The transition to 450mm wafers should typically provide an approximate 30% cost reduction, as demonstrated by scenario modeling. Further delay will only increase costs due to the steady march of the technology insertions that provide Moore s Law doubling of functionality and performance while controlling power usage. From analysis provided by IC Knowledge, there will be a significant market opportunity for 450mm (and 300mm), while maintaining both the profitability of manufacturers, and a healthy market for suppliers; and at the same time creating

18 economic benefits of a healthy growth in demand with state-of-the-art and affordable products. The European Semiconductor Equipment and Materials Industry represent a significant proportion of the World market for their products. As such, it is important that these Companies begin to invest their time and financial resources in order that they are ready to supply the semiconductor manufacturing industry with capable products when the next wafer size generation reaches high volume manufacturing. A wafer size transition also represents a good opportunity for new comers in to the market to have their products selected for use in the new manufacturing fabs. As for the 200mm to 300mm wafer size transition, and even more so for the next 450mm wafer size transition, a global cooperation between semiconductor manufacturers, equipment and materials manufacturers, and research institutions is envisaged and required to enable a timely, efficient and economically viable transition. For semiconductor manufacturers to engage with all the equipment and materials vendors necessary to fill their process development pilot lines to enable full scale technology node production, it is vital that this cooperation exists. In addition, the benefits of this development activity to smaller wafer size fabs in Europe supporting More than Moore product manufacturing may be significant. These products (discrete, analog, sensors, optoelectronics, integrated passives, etc.) typically trail behind the Moore s Law technology driver products and utilise smaller fabs. However, demand for these products is about 50% of the total capacity and does grow (albeit a slower pace) and benefits eventually from the leading edge investments. Therefore, Europe has a big part to play, and support and collaboration across all elements of the research and development operation, including National Governments is vital so as to maximize 450mm European equipment and materials manufacturer readiness and subsequent market opportunity.

19 The European Equipment and Materials (EEMI450) Initiative commenced in 2009 and offers free membership to all European based equipment and materials manufacturers supplying the semiconductor industry, that are interested in 450mm topics and activities. The goal of the Initiative is to bring common interests together, and to promote 450mm activities in Europe. By doing this, it is hoped that the competitiveness of the European semiconductor equipment and materials industry is improved, leading to an increased chance of selection by the tier 1 semiconductor companies in their future 450mm R&D and manufacturing operations. In addition, it is hoped that this Initiative can stimulate a European infrastructure that is leading in 450mm development, and as a result will induce tier 1 companies to cooperation programs which will lead to European equipped 450mm fabs in Europe. The EEMI450 Initiative has already been successful in launching the current ENIAC EEMI450 project working on 450mm proof of concepts, and also the now labeled CATRENE SOI450 and NGC450 projects in Europe. Further project proposal submissions are planned when appropriate, including the recent ENIAC Call 5 EEM450PR project; European Equipment & Materials 450mm Pilot-line Readiness.