MICRO: What have been some of the main differences, compared with your other experiences, that you've seen in getting this operation going?

MARCO MORA: I think the major difficulty is that we are really pioneers in high tech in China. When I started the 8-inch fab in Avezzano, we used not more than 50 senior engineers to start the fab and then to ramp. Here, initially, we were thinking of moving around 100, 200 overseas people. In reality, at the end, we moved almost 900 people from all around the world to start the company.

The reason we did it was because we realized the need for a strong training program, an on-the-job training program. We now have one senior engineer together with two young engineers. Just by osmosis, we transfer the know-how and we transfer the semiconductor culture to the young engineers, who are very strong because they are all from very good universities, but what they are really missing is an understanding of what our expectations are in the semiconductor manufacturing field. Also, what should the relations be between the engineer and the boss? It is not a yes-man culture, but you have to follow the procedure and be very straight on this, but at the same time, you need to be proactive, to take responsibility.

Second, when we first came here, we didn't expect to be able to follow so aggressive a plan ... initially, we were planning for one quarter, two quarters late. But when we started to work with the people, we found that these people worked very fast. There is only one small issue: They may pay a little bit less attention to details, as we are used to in other countries. And for that reason, you need to spend a lot of effort on the checks.

PHOTOS COURTESY OF SMIC

The third issue we have as a company is that there are a lot of strong people from all around the world, who came here with high expectations, and we had to learn how to work together as a team. Teamwork is a key for the success of a fast project like this. We had to understand each other, our different cultures, because even if 99% of the people are Chinese in race, they come from either the United States, Taiwan, Singapore, all around the world, so they may not understand each other very well, and we had to spend a lot of effort on that.

MICRO: Do you find with the young engineers coming up that there are any areas within the process flow where they seem to have more strength of talent; let's say, do they have a better grasp of CVD/PVD, or a better grasp of lithography or things like that, or do they have a more general background?

MORA: We tend to hire people with a bachelor's degree from the local university. The reason is because we found that the master programs and PhD programs are not really addressing what the semiconductor [industry] needs are. Now, we have started to work with Fudan University to get some theses for the students so we can fine-tune the master program and the PhD program for what we need. But this takes time.

MICRO: What about at the technician level?

MORA: We are very, very impressed with our technicians. Last month our line was around 98%, and this month, so far, we are more than 99.5%, and this is after 12 months of real production. It's really excellent, considering the fact that before we had our fab, we didn't have the possibility to do any training. We could not send our operators or our engineers to any kind of a mother fab, so we train here. We have very, very few operators with experience from Taiwan or Singapore. We are really impressed about the discipline of the operators. It is excellent.

MICRO: Was there any part of the learning curve that was more than normal, such as for cleanroom protocol?

MORA: No, because our supervisors come from Taiwan and know their job very well, and they gave the right mentality to the operators from day one. We had a bit more of a problem with the facility contractor. In the initial phase, we had the hookup people working together, and this was not an easy task.

MICRO: How are your yield teams organized? Do you have a stand-alone yield team, or is it organized across the different process modules?

MORA: We are organized like the typical Taiwanese foundry. The engineering is divided into module departments. We have around 70 engineers in the integration team for each fab, and 70% of the people in the integration team are mainly working on the product or fine-tuning a device or something like that, and they are working by project and by technology, and 30% are working on defect inspection.

We are using advanced tools for inspection inside the fab, and we have both the defect management and the yield management systems, where we can integrate all the defects and the parametric and the probing data together. On top of the integration team in each fab, there is a product engineering team, which is separated from the fab, and they act as a central focus for the yield improvement by each technology. Of course, the product engineers work using more of either the yield management system or the standard FA [failure analysis].

MICRO: How often do the teams meet?

MORA: There is a weekly meeting, during which we review all the major product runs, which includes myself, the fab director, and all the other product directors. That's where we meet with all the section managers, and we review the yield status of the major products going through the fab. Then, if there is any excursion or any problem for a specific product, we start a task force, which can meet daily.... Behind the fab, we also have the technology development team, which is separated here in memory and logic, and this team can always work as a backup in case we have a serious problem.

MICRO: Is there more logic or memory running through right now?

MORA: I would say we are still running more memory. Because here as a strategy we use the DRAM to fill the line, just to exercise or make sure we are running the maximum capacity. DRAM, you can always sell.... We are in the phase to migrate very quickly to either advanced DRAM, which when I talk is our 0.14-µm technology, or advanced logic.

MICRO: What about the other miscellaneous products?

MORA: We are running all kinds of technology here, besides flash, where we just started to do some developmental work. Mixed signal, RF, everything else...EEPROM, smart card, that sort of thing.

MICRO: Back to the yield and defect side... how is the metrology positioned in the fab?

MORA: It depends. Each technology area has its own metrology tools, but as much as we can, we try to share the tools to reduce the cost. We normally use off-line metrology tools.... But we have already started, for instance in photolithography, to move toward advanced process control monitoring. We have a closed loop of the overlay and CD, and we control the illumination and the alignment on the scanner. We do not have so much in-line monitoring system beside CMP, where we've introduced the thickness measurement.... For the advanced technology, for the poly etcher, we'll have a little bit more because we need to get better control of that machine.

MICRO: Do you see the need to have a fabwide APC system, similar to what companies like AMD have implemented at their advanced factories?

MORA: Even at TI, we started to work as much as we can [on APC], but you also have a trade-off. The good thing here is that the cost of our engineers and operators is not so high. That is why we have to push the technology where it is really needed, to make sure that our yield, our quality, is like the top-notch guys. But we may not need to do everything. We are always very conscious of cost: it is a must for a foundry.

MICRO: What's the level of automation in the fabs?

MORA: Total. About 99.9%. This is a SMIF fab, and it is fully automated.

MICRO: So it's fully automated across all three modules?

MORA: Yes, this is [a] different [approach]; it is a little bit of a new design compared with other fabs. We are the only one that has separated the front end of the line and the back end of the line. This is the first real copper fab design.... It's very separate, and the reason is because we have to deal with many different customers, and they all have their wafer protocol.... So here we can increase the capacity, even in the high-end technology, without suffering in overall output.

MICRO: Do the wafers travel from the FEOL on an automated system into Fab 3B?

MORA: Yes. They are also connected to Fab 1 and Fab 2, which are also connected, and each fab sees the other fab as their virtual fab, as a backup.... We designed this fab to run copper because of cross-contamination issues. We accommodate a lot of issues for the wafer movement protocol inside that design.

MICRO: Is the CMP portion of the fab also somewhat segregated, or is it all just merged now?

MORA: All our CMP are dry in and dry out, so there is no big issue. Even for the implanters.... One of the characteristics of this fab is that all the process tools are on the same floor to minimize the automation. We do not use any lifters for automation.