Shang-Yi Chiang, Senior Vice President of R&D at TSMC addressed several hot discussed issues at the TSMC Japan Executive Forum in Yokohama earlier this week, including the company’s 40nm capacity & yield issues, high-k and 28nm & 22nm technologies.

1. 40nm capacity issues

According to Chiang, the 40nm technology had a very high demand in the early stage, higher than any earlier generations of manufacturing process.

At present the Fab 12 is capable of manufacturing 80,000 pieces of 12-inch wafers per quarter, and after the Fab 14 is also ready for tape production of 40nm, the capacity will be doubled by the end of this year, reaching 160,000 wafers.

2. 40nm yield issues

Chiang claims the 40nm yield problem was resolved in the second half of last year, and the company is now building the capacity very aggressively to fulfill the high demand from customers.

He also notes the transition to 45 and 40nm is more challenging than ever. This is the first time they began to use 193nm shrink immersion, so there is a very high potential defect. TSMC began to use the second generation low-k material with a k value of 2.5 which can become quite fragile, so moving to 40nm got pretty challenging.

3. Process at 28nm

The first node TSMC is going to release for the 28nm will be known as 28 LP – the poly gate and silicon oxide nitrate version. The mass production is expected to start at the end of June this year.

4. Process on high-k and metal gate

28 HP, the first high-k metal gate for the high performance application will be introduced at the end of September this year, and three months later we’ll have the 28 HPL.

5. Plans for 22nm node

TSMC plans to introduce 22nm node about two years after 28nm, so it would happen around the Q3 of 2012. The original version would be designed for high performance applications and then the low power version at the end of Q1 2013.

6. Litho directions

TSMC will continue to use the 193 immersion with double patterning at first, and hopefully migrate to EUV or multiple e-beam direct write when they become more mature and more cost-effective.

Of course the company need pay lots of money for the new technology – one single EUV tool with a matched track costs as much as 80 million dollars.