by Bob Haavind, Editorial Director, Solid State Technology
When the next upturn comes, chipmakers aim to have much leaner, more agile fab operations ready to capitalize on it. That was a dominant theme at the International SEMATECH Manufacturing Initiative (ISMI) conference in Austin, TX (Oct. 22-23). Chip factories will also run greener, cutting costs at the same time, by using less energy, water, chemicals, and other consumables.
The theme was set in the opening keynote by Masakazu Kakumu, Toshiba group VP and VP of Toshiba Semiconductor’s system LSI division, as he laid out Toshiba’s strategic vision for the next couple of years.
While growth has slowed in 2008, Kakumu said, and the company’s suppliers have been “hard-hit,” he expects a strong recovery in 2010. The company’s goal is to reach revenues of ¥2T (US ~$20.5B) with 10% operating profits that year, while retaining its No.3 share in global semiconductor revenues. The strategy aims at enhancing profitability of existing businesses while incubating new growth engines, he explained.
Masakazu Kakumu, corporate VP, Toshiba; VP of system LSI division, Semiconductor Co.
A major part of that effort will be boosting productivity and maximizing output of both 300mm and 200mm factories. While total capex will dip to ¥367B ($3.76B) in 2008 from ¥411B ($4.22B) last year, memory capex has remained about steady, a little over ¥50B ($513.3M) over the past three years. He also said that for NAND memory, two new state-of-the-art fabs (Lines 5 and 6) will be built simultaneously at Yokkaichi. Toshiba is also purchasing equipment from Sony’s system LSI operations, he added.
Toshiba will focus on next-generation factory (NGF) methods to boost productivity, improve yields, and gain savings through environmental programs, to achieve more energy efficiency while cutting PFC and CO2 emissions, for example. Kakumu showed a calculation to indicate how all these programs combined should double Toshiba’s investment efficiency by 2010.
Technology development costs will be cut by alliances, collaborating with consortia (such as ISMI), and national programs, he said.
Kakumu also said Toshiba aimed to maintain its No.1 global position in discretes, but also wants to push it into double-digit market share by expanding in areas such as power and small signal devices. Current output of 10k wafers/month should grow to 60k wafers/month by 2010, he predicted.
No waste of energy
Efforts to save energy and make other environmental improvements must be analyzed on a fab-wide basis, not just for individual tools, suggested Andreas Neuber of Applied Materials. To aid this analysis, he described a modeling tool for eco-efficiency engineering that helps decision-making at the fab level.
Energy used by process tools is becoming more important, Toshiba’s Kakumu pointed out, because while it is 47% of the total for 200mm fabs, it goes up to 54% for 300mm fabs.
Neuber described a “Fab-Eco Impact” model, and then gave some examples of how it might be applied to areas such as energy and chemicals. The model identifies phases for engineering studies, starting with developing rough benchmarks that can be applied to identify impacts across all fab operations. This is followed by an idea/innovation phase to identify opportunities for improvements. The analysis then tracks through fab systems to spot potential bottlenecks.
The model must be used for both operating and idling flows, and it shows that there is significant energy usage during idling, for example. At 100% capacity, a fab still has some 25% idling time, and if utilization drops to 70%, idling may rise to 50%. This offers great opportunities for savings, Neuber said. Chillers, for example, are used to cool water to 17-22°C before feeding it to tools — but many tools can operate with water at 32°C, so it could be fed directly from the cooling tower without a chiller.
Neuber also described a smart idle mode for tools in which pump speeds are reduced and less water is fed to the tool, for example. These variations are then tracked through all flows in the fab by using the model. He also cited a case study on a dilution purge that showed savings of nitrogen, water, natural gas, and power across the fab.
Mapping NGF progress
Many emerging concepts for improving fab productivity, yields, and throughput are being incorporated into the NGF under a 19-point guidelines approach developed through ISMI, according to Scott Kramer, VP of manufacturing technology. The NGF targets for 2012 aim to enable new fabs to be built with cycle times reduced 50% and costs down 30%. While the NGF concept is being aimed at new fab design, priority within the 19-point guidelines is being given to methods that also can be retrofitted to existing fabs.
A five-year NGF Roadmap will help identify future engineering projects needed to meet targets. Kramer indicated that ISMI will increase its collaboration with process toolmakers, and the five-year Roadmap will be adjusted to align with vendor roadmaps.
Tom Jefferson, an Intel assignee to ISMI, described an energy-reduction vision for NGF, for example. A 450mm wafer fab built along NGF guidelines would have no per-wafer increase in processing energy. Achieving this would require a 55% reduction from the current 2.13kWh/cm2 to 0.95kWh/cm2.
The focus of the NGF program is waste minimization, according to Kramer, and a major effort is going into tools for waste identification using visualization and analysis. One lesson learned from the 300mm transition, according to Kramer, is that it is essential to consider the needs of high-mix, low-volume fabs as well as those of high-volume chipmakers. He also pointed out that ISMI develops continuous improvement programs for 200mm and 150mm fabs.
Lean manufacturing approaches
Lean manufacturing principles were also widely discussed at the conference to help reduce waste and boost productivity. Steve James, Dallas wafer fab (DFAB) manager for Texas Instruments, testified to the dramatic impact of such methods on fab operations within TI. One simple program dubbed “5S,” came to TI’s attention when its Hiji, Japan, fab was ready to be shut down due to poor results. The 5S program, a disciplined, orderly process for identifying and cutting waste, was introduced and provided rapid improvement. It emphasizes orderliness such as no junk lying around and clear labeling throughout the workplace, so it is obvious what needs to be done without referring to manuals or instructions. The program requires audits, daily and in detail for team members, up to monthly with more overview for upper managers. One goal James cited for the process was reducing human error by half each year, which has been met.
Steve James, Dallas wafer fab manager, Texas Instruments
James also described Kaizen Blitz events as a method to implement lean manufacturing. (Kaizen means continuous improvement.) These involve 10-15 people across multiple functions who apply the lean principles to their own tasks and worksites, and are empowered to make improvements themselves. Last year, TI’s DFAB conducted 46 Kaizen Blitz events involving 455 people, and productivity has been rising steadily.
Examples of “extended 5S” in DFAB.
CLICK HERE to view larger image
Another principle James extolled involves “AND” thinking. The idea is to take opposite goals, such as flexibility or continuity — then “or” is replaced by “and” as work teams find ways to achieve both, rather than one or the other.
A lean manufacturing approach to solving big problems that might take months to identify and analyze was presented by Ron Warren of IBM, Burlington, VT. Even if a solution is developed, implementing it may prove extremely difficult. Asking 100 engineers to do their job differently may be a big challenge, Warren suggested.
To achieve change like this, Toyota studied job instruction training methods devised in the US in the 1910-1920 era, he said. A job breakdown sheet is developed by team members, and a four-step training method is applied by certified trainers to get best methods applied uniformly by everyone involved.
The follow-up phase proved to be the most difficult, Warren said, because many workers stray from best practices but resent someone looking over their shoulder to find flaws. Instead, a better approach has been to engage the workers in the follow-up, finding out how well the procedures are working, and whether more improvements can be made. This has improved the implementation process, and in 2007 IBM trained over 700 employees and 12 new trainers were certified.
Whenever the turnaround does come, 2010 or earlier, efforts like these should help ensure semiconductor factories will be geared up to take full advantage of it. — B.H.