by Ken Goldstein, Ph.D.
There is a tremendous amount of discussion of late regarding energy conservation in the operation of semiconductor facilities. It's my contention that the energy saved by this approach is relatively unimportant in the overall scheme of things. Of course, no one wishes to waste energy needlessly, but reducing energy consumption solely for the sake of energy conservation should be well down our list.
 |
As might be expected, fabs represent very expensive real estate. Commonly used capital estimates can range from $1,800 to as much as $4,000 per square foot of process cleanroom space. That's a significant sum that includes facilities capital only. It buys the cleanroom (building shell, cleanroom walls, floors and ceilings), and the supporting utility systems (high purity gases and liquids, effluent treatment). We must add the cost of the process tools, which can range from $1.5 million to $8 million apiece, and will constitute up to 80 percent of the cost of the facility. A new state-of-the-art 300-mm fab is likely to cost in the neighborhood of $2.0 billion—$1.5 billion for the tools and $0.5 billion for the facility
These manufacturing plants are expected to run around-the-clock, without interruption. Planned shutdowns are typically less than one day per year for required maintenance. This around-the-clock mode of operation consumes significant electrical power. Consequently, these facilities ought to be likely candidates for energy-conservation measures. This may not be the case.
What is the purpose for these costly facilities? In light of the costs just described, it would be fair to suggest that the facility's only role is to maintain the environmental conditions and provide the ultra-high purity utilities required for production, all with near perfect reliability. Everything else, except safety, is secondary.
With those elements in mind, the key attributes of such a facility are that it be: safe; non-polluting; clean (high yields); productive (optimization of tools and layout); reliable (around-the-clock) and flexible (for future tool sets and product families).
Energy conservation should be well down this list. Yet, one may argue, money is money and if we could significantly reduce our facility electrical consumption would we not significantly reduce our manufacturing costs? Yes, that would be a great thing to do, but we can't get there.
Here's why. As reported in an article by M.C. Williamson of the EPA (Semiconductor Fabtech, eighth edition), the energy consumption measured in 12 semiconductor facilities breaks down as follows:
Air 19 percent
Chilled water 20 percent
Support/lighting 3 percent
Exhaust 7 percent
Tools 35 percent
DI water 5 percent
Process cooling 4 percent
Nitrogen 7 percent
A review of these numbers reveals that less than half (49 percent) of the total energy consumed in a typical fab appears to be facility-related—air, chilled water, exhaust and lighting. The remainder is process-related—tools, process gases and process liquids. Let's leave the process-related components to the tool manufacturers to work out and examine the facility components for their energy savings potential.
Support & lighting (3 percent): Other than walking around in the dark, there is not much to gain here.
Exhaust (7 percent): Although this component appears to be facility-related, it is actually driven by process needs (heat exhaust, corrosive exhaust and solvent exhaust) as well as safety concerns. We can achieve some savings here by using high-efficiency motors and by ensuring that our exhaust systems match our process and safety needs without any additional overkill.
Chiller (20 percent): Here is a large component in which we can really sink our teeth into; except that about 60 percent of the heat load comes from the process tools. The remainder is determined by climate, by recirculation fan heat and by the temperature and humidity requirements of our cleanroom.
Air (19 percent): Our design team knows their stuff, so they have already minimized the area under filter and designed a low pressure-drop system. All we have left to fiddle with is our airflow velocity. If we reduce this from 90fpm to 70fpm we will reduce our fan energy consumption by about 20 percent, or about 4 percent of the total energy consumed. By doing this we also reduce the chilled water load somewhat since we will be producing less fan heat. However, we have altered the contamination-control properties of our cleanroom. This may work, but it may also have some risk involved.
When all these measures are taken we may find that we have reduced our facility-related energy costs by 10 percent. But, we may have altered some process-required parameters in order to do so—parameters that relate directly or indirectly to product yield.
Although energy costs can be large, they typically represent only five percent of the cost of the wafer. Close to 75 percent of the power that's consumed is dictated by the process tools. This leaves approximately 25 percent of the total energy consumed that can be significantly influenced by the design of the facility. If we were to set a goal of reducing this consumption by 10 percent, the net impact on the cost of the wafer would be in the neighborhood of one-tenth of one percent.
By experimenting with the numbers one may find that the savings to be realized as a result of our facility energy conservation efforts would not pay for: One industrial accident; one environmental incident; one half day of downtime; one half of one percent product yield, or; one miserable tool replacement that could have been much easier if we had only planned better.
Perhaps our efforts would be better spent addressing these issues.
Ken Goldstein is principal of Cleanroom Consultants, Inc, in Phoenix, Arizona, and is a member of the CleanRooms Editorial Advisory Board. He can be reached at [email protected].