IC: The trend of 450mm silicon wafers in the 22nm era

The Semicon West Show in California in July each year is one of the largest semiconductor equipment and materials exhibitions in the world. Since the exhibition was held in July, it was just the first half of the year. Therefore, during the conference, many senior executives expressed their opinions on the development and prospects of the industry. This article attempts to summarize the progress of the exhibition and the industry in the first half of the year to discuss topics that are of particular concern to the industry.

The growth of the semiconductor industry is not suspense The changes in the semiconductor industry have been declining in the first two quarters of the previous year. However, there should be no suspense in maintaining growth. How much growth depends on future market demand.

The global semiconductor industry is changing like a drama. 2010 was the first high growth year after the international financial crisis, with an increase of 32%. Just entering 2011, the industry first rationally believed that this year's growth would not be possible, but driven by inertia and the end market for electronic products such as smartphones and tablet PCs, the analysis institutions at the beginning of the year basically The semiconductor industry in 2011 is expected to grow between 2% and 10%.

With the past two quarters, there has been a slight change in the situation. In addition to the impact of the strong earthquake in Japan in March, it is mainly hampered by the global economic environment, such as the slow recovery of the US economy and the high unemployment rate. The debt crisis has not subsided. According to IHS iSuppli's latest research data, semiconductor suppliers' semiconductor inventory levels rose for the seventh consecutive month in the second quarter as the industry rebuilt stocks and prepared for expected growth.

Normal Q3 is a traditional peak season every year. Today, due to global economic difficulties leading to a decline in consumer confidence, there may be an anomaly in the peak season, so many market analysts and companies have recently lowered their expectations for semiconductor growth this year. For example, Carnegie’s Bruce Diesen forecasted a growth rate of 5% in May, down from 3% in July; Gartner forecasted a growth rate of 6.2% in Q1, and downgraded to 5.1% in June. Only IHS iSuppli Corporation forecasted a growth rate of 7.0% in April and an increase of 7.2% in June.

The above only reflects the past two quarters in the past, the semi-conductor industry's changes to the direction of weakening, but the overall industry fundamentals are still normal, such as this year's investment in semiconductor equipment from last year’s 38.5 billion US dollars to 44.3 billion US dollars this year, an increase of 12% . The terminal electronic product market is still quite dynamic. For example, in 2011, the market growth rate of consumer electronics products will reach 5.6%, which is higher than the US GDP growth rate of 2.4%. The revenue of consumer electronics products will continue to climb in 2012 and is expected to reach US$197 billion. In 2011, the sales volume of tablet PCs, including the Apple iPad and its rivals, will reach 26.5 million units, which will generate revenue of 14 billion US dollars for various device manufacturers; sales of smart phones will increase by 45 percent over the same period of 2010. %, up to 23 billion U.S. dollars.

In current consumer electronics, sales of flat-panel TVs have a downward trend. Research shows that 88% of American households have at least one digital television. It is because of such a high penetration rate that TV sales will decline in 2011, but revenue will still exceed 18 billion US dollars.

IDC recently announced its latest forecast that the semiconductor market will grow from 282 billion US dollars in 2010 to 303 billion US dollars in 2011, an increase of 7.4%, and it is predicted that it will increase by 5% in 2012 to reach 318 billion U.S. dollars during 2010-2015 The average annual growth rate of semiconductors is 6%. According to the classification, the CAGR of the computer IC is 4% to 5%, that of the communication is 7%, and that of the consumer is 5%.

Therefore, the growth of the global semiconductor industry this year should be no suspense, and how much it needs to increase depends on future market demand.

Verifying that the acquisition of Hengda's laws by the big players makes this market appear a "giant" and that the semiconductor industry's "Greater Evergrande" law will once again be verified.

“Across the history of the chip industry, companies that have seized the opportunity at the right time have achieved very good growth.” In May of this year Texas Instruments acquired US National Semiconductor (NS) for $6.5 billion at a high price for everyone’s Eyes light up. This transaction is not only the highest amount of M&A in TI's history, but also the highest amount and largest chip industry M&A transaction since 2006. Coincidentally, almost the same time, TI's competitor Infineon also acquired Qimonda's 12-inch fab in Germany for 100.6 billion euros. The issue of "scaled survival" of chip makers was again raised.

According to Bloomberg data, in the past year, there were 196 M&A transactions in the semiconductor industry. The average size of these transactions was US$ 129.8 million. In contrast, the difference paid by TI is more than four times the average. The industry commentary TI's acquisition of NS undoubtedly pushed this round of expansion to its peak.

According to Databeans' report, the global analog semiconductor market in 2010 was US$42 billion. TI's revenue in this product line reached 5.98 billion U.S. dollars, accounting for 14% of the global market. In 2010, NS revenue was approximately US$1.6 billion, with a market share of 3%.

For a long time, TI and NS have always been competitors, and each company has its own unique competitive advantage. TI has 30,000 analog products, extensive customer influence, and industry-leading manufacturing technologies, including the world's first 12-inch analog chip manufacturing facility. NS has 12,000 kinds of analog products, occupying a strong position in the industrial market in particular. With the 2010 sales combined, TI will replace Toshiba as the third largest semiconductor company in the world.

For chip manufacturers, especially foundries, "scale" is not a new concept. Over the past two years, the semiconductor industry has seen a trend of such vertical integration. Japanese chip giant NEC merged with Renesas Technology, Abu Dhabi ATIC and AMD formed a joint venture Global Foundries. In January 2010, Global Foundries completed the merger of Singapore Chartered Semiconductor.

However, large-scale mergers and acquisitions rarely occur in the analog chip market. Some industry insiders believe that mergers and acquisitions make this market appear "giant", the semiconductor industry "Greater Evergrande" law will once again be verified.

"There are three important aspects of staying ahead of the current market." TI's CEO Templeton said, "Product technology, low-cost rapid manufacturing capabilities, and a global technical service system, some companies may only be very strong in some areas. However, the lack of these three conditions is also the reason why TI can continue to develop."

The process technology entering the 22nm node semiconductor process technology is still the progress of Moeller's Law. In 2011, it can enter the 22nm node, and the time to enter the next node is still difficult to predict.

Intel has announced its roadmap. Two years later, Intel will produce chips at 14nm, and then in 2015, it will produce chips at 10nm. Finally, Intel plans to produce its first 7nm chip in 2017.

Intel's 22nm tri-gate transistor technology is a milestone, indicating that semiconductor technology will transition to 3D.

According to the ITRS semiconductor process roadmap, at the 45-nm node in 2007, Intel released a high-k/metal gate technology that can be seen as an innovation in the composition of transistor materials, replacing high-k materials with traditional SiO2, and letting the law Extended for 10 to 15 years. In May of this year, Intel released a 3D transistor structure, which changed the two-dimensional structure of traditional transistors into three-dimensional. It should be another major revolution in semiconductor process technology.

There are also many speculations about the technology nodes that Intel will adopt. Intel's 22nm process will be based on Intel's third-generation high-k/metal gate approach, which uses copper interconnects and low-k technology. Like 32nm, Intel uses 193nm immersion lithography.

Intel's 22nm 3D transistor technology, performance 40% higher than 40nm, 30% power consumption, the cost of its technology increased by only 2% to 3%, while the use of SOI technology to rise 10%. In addition, 100% battery life can be achieved, and trial production is expected to begin in the first half of 2011.

According to Intel's latest report, the 22nm product was put into production this year. The 14nm plant is increasing investment and stepping up construction. It will soon receive orders for 14nm devices. In the future, semiconductor technology can reach 7nm nodes.

Taiwan Semiconductor Manufacturing Co., Ltd. will mass-produce on the 28nm process during the year and will officially begin production of wafers based on the 28nm process at the end of this year. TSMC said that the company plans to start commercial production of the 28nm manufacturing process sometime in Q3 2011, and by 2011 Q4, the 28% wafer-based revenue contribution rate will reach 2% to 3%.

Compared with the original upgrade to the 40nm manufacturing process, TSMC's upgrade of the 28nm manufacturing process will be smoother both in terms of increased production capacity and improvement in yield, because TSMC would need to upgrade its equipment when upgrading its 40nm manufacturing process. The upgrading of TSMC's 28nm manufacturing process seems to have completed the commissioning of new equipment.

Samsung began mass production of 27nm NAND Flash in April 2010, opening the era of global 20nm-level manufacturing process. After 15 months, Samsung has developed to 21nm products to demonstrate its unparalleled global technological level.

In addition, Toshiba will mass-produce NAND Flash in 19-nanometer process at the fab 5 semiconductor plant in Yokkaichi, Mie Prefecture, Japan at the end of July this year. Toshiba currently mass-produces 24nm process products. After Hynix is ​​not in the technological competition, it will mass-produce NAND Flash in 20nm process in the fourth quarter of 2011.

EUV is the inevitable trend of future lithography In addition to EUV technology, there are many candidate technologies such as unmasked multi-beam electron lithography and nano-printing, and the future progress depends on the cost performance.

One of the two wheels driving the progress of the semiconductor industry is the shrinkage of feature sizes and the increase in the diameter of silicon wafers. Needless to say, size reduction always prevails. Intel recently announced the successful development of the 22nm 3D technology, indicating that the 16nm technology with unclear prospects may be introduced into the market in advance.

Lithography, which affects size reduction, is currently using shorter exposure wavelengths, such as ArF light sources, with an oscillation wavelength of 193 nm. Due to the many postponements of EUV (extreme ultraviolet lithography) technology, the 193 nm optical lithography method has been brought into full play. From 193nm dry process to wet process, up to the double graphics technology of process complexity and cost increase.

As early as 2008, the industry thought that optical lithography was nearing its end, and 22nm was the end point. After that, EUV technology with a wavelength of 13.4nm must be adopted. Although EUV may be a candidate for next-generation lithography, EUV has been postponed several times due to unresolved issues, including the lack of a suitable light source (not enough power) and pellicle for EUV masks. There are also economic factors, such as the price of up to 125 million U.S. dollars per device, as well as the hourly silicon output of the device.

In addition to EUV technology, there are many candidate technologies such as maskless multi-beam lithography and nano-printing. The future progress depends on which technology is more cost-effective.

The current progress of global EUV and maskless electron beam lithography equipment is as follows:

Nikon proposed its latest optical system, the 621D, with an overlay accuracy of 2nm/3sigma. The NA of its development EUV-1 is 0.25, and the NA is 0.4 when the production EUV-HVM is introduced in the future.

ASML has set its EUV roadmap with a NA of 0.25 for the development NXE3100. If the light source energy reaches 15 mJ/cm2 on the surface of the wafer, it can reach 125 pieces per hour. ASML currently has three devices tested on the client, the fourth one is being installed, and it is predicted that 10 units can be shipped by the end of 2012. As the global supplier of EUV light sources includes Cymer, Gigaphoton and Ushiro, there is also the emerging plant XTtreme. ASML has not made a final decision on its selection of EUV light source suppliers.

Another Mapper company is developing a maskless multi-beam electron lithography machine. Its first generation machine 1.1 produced 1 tablet per hour at 25nm and was shipped to France's research and development agency Leti, which will ship it to TSMC next year. The company plans to use bundled electron beams to increase production capacity to 10 pieces per hour and eventually achieve 100 chips per hour at 16nm. Mapper's strength lies in the price and capacity of each of its equipment. The goal is to have 5 million euros per unit, which is only fractional compared to EUV prices.

Another EUV light source supplier, Xtreme, uses laser-discharge plasma technology. The light source device consists of two wheels containing liquid tin, which are discharged at 10,000 Hz to generate laser pulses (up to 100,000 Hz in future production facilities). The weight of the light source can weigh up to 8 tons. In order to prevent tin contamination, a shielding cover is installed around the condenser. Xtreme's Marc Corthout believes that the light source actually achieves the power required for the process requirements.

IMEC of Belgium believes that EUV is an inevitable trend for future lithography. Starting from the needs of the global Internet, it will certainly prompt logical or storage customers to adopt it. The three key elements of EUV are devices, masks, and photoresists, which have now been confirmed to be practicable. Despite this, the main challenge of EUV is still the light source, and it is expected that the real adoption may be introduced in the 11nm process.

IMEC introduced the exposure equipment stand-alone and EUV light source in March 2011, and began adjustment and integration testing within the research facility. It is now connected to TEL's coating development device "Lithius Pro". IMEC said: "The throughput is 20 times that of the EUV exposure device used for ASML's exposure process evaluation. It plans to achieve 100W of light source output power and 60 sheets per hour of throughput in early 2012." In terms of overlay accuracy, this achievement has The potential ability to achieve a target value of 4nm or less. The resolution can reach below 20nm when using dipole illumination. This achievement is a milestone for IMEC's ​​goal of establishing a mass production technology with a resolution of 16nm in 2013.

The industry is still doubtful about the prospects of EUV, and next year is the key. Despite this, EUV has greatly improved compared to a year ago, and there may be two devices that will enter the preproduction stage.

It is certain that the use of traditional optical lithography equipment is the most economical in the industry, but there is always an end to it. Especially when the size is getting smaller and smaller, the number of manufacturers willing to follow because of the high cost will be less and less, which will lead to poor economic feasibility in the future. Therefore, some people questioned whether EUV can be introduced at 11nm.

The key to the success of 450mm silicon wafers is to promote the rapid transition cost reduction of 450mm silicon wafers. Only by enabling both chip manufacturers and equipment manufacturers to implement dual-entry at the same time, can they continue to progress and develop.

The 450mm wafer transition has been debated in the industry. The three companies that have taken a positive attitude so far today are Intel, TSMC and Samsung. TSMC’s attitude seems to be more clear. For instance, Qin Yongpei, Vice President of TSMC's Operations and Product Development has talked about the 18-inch fab plan, and reiterated that TSMC will set up a trial production line from 2013 to 2014, and mass production at the Taichung Plant from 2015 to 2016; 18-inch wafer production will be Environmentally and economically, it will be more efficient than a 12-inch plant.

Another view held by the industry on the transition of 18-inch and 450-mm silicon wafers is a group of equipment suppliers headed by Applied Materials. The reason is also very clear. The cost of developing 450mm equipment is about US$20 billion. Who will bear it? On the other hand, it is more practical, that is, how many orders can be used in the future to increase the return on investment.

However, due to the recent progress in the industry, the views of both sides have become closer and closer. For instance, Applied Materials' President Mike Splinter has publicly stated in his Q2 summary meeting this year that in 2012 the company will increase the investment in research and development of 450mm equipment, indicating that Applied Materials has seen 450mm wafers as the trend of the times and as global semiconductor equipment. The leader can no longer have the slightest hesitation. Only active follow-up is the way to survive in the future.

Nowadays, doubters are rare for the transition to the 450mm wafer. However, there are still different opinions as to who is the real first mover and at what time.

Because the ideal evolution process was to start a 450mm trial production line with two or three companies at the beginning, the transition to mass production lines quickly occurred due to the decrease in chip manufacturing costs, which is almost the same as the previous transition from 200mm to 300mm silicon wafers. . Due to the same process conditions, the operating cost of the 450mm production line is only increased by 30% compared with 300mm, but due to the 2.25 times increase in silicon area, the manufacturing cost of the final chip is reduced, which will stimulate the expansion of production capacity and more. The plant put 450mm wafers (estimated more than 10 worldwide). Such a process will lead to a market share of 450mm silicon wafers will be small to large, such as the current 300mm silicon wafers have accounted for more than 60% of total silicon shipments. Therefore, the key to the transition to 450mm wafers is the cost reduction, and it must also enable chip manufacturers and equipment manufacturers to achieve a win-win situation.

As for the transitional time point of 450mm silicon wafers, TSMC chose to have a mass production stage from 2015 to 2016, ie, 22nm to 16nm, which may have different opinions in the industry. Because in the transition from 200mm to 300mm wafers, originally estimated at the 250nm node, actually postponed to the 130nm node, Intel first built a 12-inch production line in 2002. Therefore, the supporting industry chain of 450mm silicon wafers will become a constraint factor in the future, resulting in a different view of the industry is completely normal.

From the perspective of Applied Materials, it has been developing 300mm equipment since 1996. It was originally thought that there would be a large number of orders in the year 2000 or so. In fact, due to the Internet bubble in 2001, the actual global 300mm equipment was delayed until 2003, so the company’s profit was postponed. In 4 years, such historical lessons cannot be completely forgotten when transitioning to 450mm.

Mik Splinter, president of Applied Materials, recently answered an analyst's question about the 450mm equipment process. He believes there is still a lot of work to be done, such as equipment automation and cavity design, but in 2012 the company will definitely increase investment. However, since it is not yet clear when the customer will place an order, it is still not easy to provide a prototype. However, some manufacturers are considering mass production of 450mm wafers from 2015 to 2017.

The cost reduction is the key to the success of the 450mm wafer. However, this refers to the manufacturing cost of the chip. It is not only related to the equipment, but also related to the supporting industrial chain. It is believed that only if chip manufacturers and equipment manufacturers achieve a win-win situation at the same time can they continue to progress and develop. Therefore, the transition to 450mm wafers in the future is estimated to be more complicated and difficult than the transition to 300mm silicon wafers, and the cycle may be longer.

In general, 450mm wafers are the trend of the times and the industry has a consensus. However, it is still to be seen when the real start of the transition and how many manufacturers around the world are willing to invest from US$7 billion to US$10 billion to build factories. This is also the crux of what seems to be different views, because only the sales may reach or approach 200 A billion-dollar company can afford such a huge amount of continuous investment.

Magnetic Powder Core

Anyang Kayo Amorphous Technology Co.,Ltd. , https://www.kayoamotech.com

This entry was posted in on