In the late 1980s and early 1990s, tremendous yield gains were realized through the implementation of ultra-high-purity (UHP) systems that helped produce devices with cutting-edge geometries. With the push toward submicron geometries and target yields in excess of 90%, semiconductor fabs worldwide began to embrace contamination control during the construction and start-up process, employing virtually any technology or protocol that presented even the potential for minimizing contamination and improving yields.

These contamination control strategies increased fabs' ability to manufacture submicron devices. Cleaner gases and liquids meant that the critical purity thresholds for killer defects could be surpassed. Given the gains that most fabs were realizing in the late 1980s and early 1990s because of investments in higher-grade materials and enhanced QA/QC procedures, it was easy to justify elaborate quality control or production methodologies. It simply made sense commercially to accept the added labor costs associated with extensive QA/QC protocols and the capitalization burden of piping fabrication cleanrooms because these costs were offset by chipmakers' heightened ability to manufacture increasing quantities of smaller devices.

Figure 1: Randomly sized holes cut in plastic caps used in place of standardized outlet fittings for purge restrictors during orbital welding. The lack of standardized outlet fittings with uniform hole sizes or differential pressure gauges leads to avoidable variations in weld bead profiles.

The Horizon of UHP Technique

While the construction QA upgrades of the past made significant yield increases relatively easy to accomplish, further gains from the protocols of the late 1980s and early 1990s are now difficult to realize, or incremental at best. The principles of contamination control for building fabs that process submicron ICs apply to the construction and installation of process piping, but semiconductor devices and piping are not equally vulnerable to absolute compromise from contamination. Many of the elaborate construction techniques used to protect wafers have not proven beneficial to UHP systems. Consequently, some of them have been abandoned, such as the use of stainless-steel purge lines for tube bending rather than flexible PFA and awkward HEPA-fed mobile glovebags for field-piping assembly and fit-up.

Efforts to maximize profits have increasingly emphasized cost savings over contamination control. Thus, the traditional paradigm of maintaining large cleanroom environments at ultraclean levels has given way to the use of ultraclean minienvironments such as SMIFs. The physical science of contamination control still applies in the semiconductor industry for UHP systems, but it is being refined to be more effective and commercially viable.

Once most of the major players had instituted state-of-the art UHP and the point of diminishing returns had been reached, facilities managers initiated measures to minimize start-up costs, reduce unscheduled downtime, capture market share, and expand into foreign regions with low labor costs. Foreign partnerships gave U.S. and other big IC manufacturers access to protected markets in developing regions.

If projects in developing countries had the time and money to proceed with elaborate, laborious protocols for prefabricating, installing, and starting up UHP systems, they could construct clean systems that would come on-line quicker and be easier to troubleshoot in the pilot-line and ramp-up phases than systems that do not follow these protocols. That sounds great, but it does not work that way. Developing countries, by definition, are still developing and, therefore, are not fully prepared to employ the most cutting-edge methods to construct fabs that will run £0.25-µm processes on >200-mm substrates.

Figure 2: Convexity and concavity in a UHP piping system resulting from variations in weld ID purge pressure.

Even in countries with a developed semiconductor industrial base, where mechanical contractors are used to the drill of cutting-edge UHP methodologies, QA authorities still must fight to achieve total compliance with the construction protocols of first-tier semiconductor fabs. Installation contractors always resist such protocols as a matter of doing business; protocols cost money and reduce profit margins. They accept such protocols, knowing that they will haggle with QA to minimize expenditures, maximize the company's profit margins, and increase their personal bonuses. Installation contractors in developing regions are not in any better position and also must contend with the commercial pressure of making a profit while lacking fully developed UHP industry support infrastructures.

In Southeast Asia, for example, an experienced and salty QA consultant knows the limitations of a project before arriving on site and has planned for the inevitable lack of infrastructure, equipment, consumables, and skilled labor. The consultant also has a pragmatic attitude toward schedules that are based on a set of theoretical assumptions made by office personnel back home.

Effective consultants not only identify problems, but also provide solutions. That may be a significant departure from the typically static role of conventional QA, but experience in Southeast Asia has proven that QA consultants must perform pipefitting instruction, contamination control training, one-on-one coaching of welders, or even hands-on repairs of equipment in order to keep things going and maintain quality.

Figure 3: Construction site at which proper materials such as consumables for contamination control are lacking. The purge fitting is old and dirty, there is no point-of-use filter on the purge line, and cylinder gas is being used instead of a cryogenic Dewar source.

It doesn't do any good to cry that contractors knew what they were doing when they bid to the specs and now must make good on their commitments. The hard truth of most jobs in the developing world is that piping contractors cannot comply completely with existing specs, but the facility still must be built, even if under less than optimal circumstances. Therefore, a premium is placed on the experience of hands-on field engineers with the breadth of knowledge to know which corners can be cut without jeopardizing fab safety or the semiconductor process itself.

Uncompromising, blind adherence to a theoretical specification, designed for execution in a perfect world, is impossible in developing countries. And technicians or engineers without the flexibility or depth of knowledge to provide creative solutions or work around problems do little to serve their clients. It is advantageous to use third-party QA entities if the solutions that they engineer allow projects to meet commercial commitments while preserving the intent of the specs.

Silicon doesn't know if it is in the United States, Europe, or Asia. The chip-manufacturing process is constrained by the laws of physics and chemistry. But if the potential is low that a protocol compromise or methodology change will jeopardize chip processing, nothing will be gained by halting work because an installing contractor is not complying with construction protocols.

A significant safety margin is designed into most specifications and fabrication procedures purposely. In the salad days of the late 1980s and early 1990s, fabs threw incredible amounts of money at problems and developed many hard-to-discard safety margins. But when faced with inexperienced contractors or inherently limited infrastructures, many cherished safety margins must be sidelined because of the inescapable realities of the project at hand.

QA reps once were considered to be the client's advocate, defenders of the faith protecting the client from overzealous or unscrupulous contractors. This view is at least partially true, because QA provides a measure of project control. But nowadays, QA is widely held to be a necessary evil at best and a source of aggravation that constantly must be throttled to prevent it from handicapping the schedule unnecessarily.

In actuality, QA is fundamentally constrained by the customer specifications, which are written to completely limit risk and incorporate every conceivable measure to guarantee the fab as great a safety margin as possible for ensuring system purity. But in the real world, a certain degree of risk is inherent in developing areas, where local contractors cannot comply to the letter with customers' specs in a timely manner. A calculated balance must be reached by eliminating superfluous requirements from the safety margin to allow the job to proceed, while preserving the integrity of the systems.

Figure 4: A typical field purge setup for welding mains on which the coaxial tees and terminator fittings are handmade and point-of-use filters are lacking because they are cost-prohibitive.

Such decisions cannot be made unilaterally by QA, but must be the result of objective debate between the client and the QA consultant. Project specifications can be dispensed with only if the client approves. QA entities are contractually bound to support project specifications and cannot grant exceptions independently without risking litigation. However, they may offer technical reasons why a project specification should be modified. If the client feels that the modification will jeopardize the chipmaking process, he knowingly accepts the delays associated with strict conformity to construction specs. If the technical grounds for modifying specs are supportable, the client can agree to modify them on the basis of facts.

Most seasoned project managers realize that establishing foreign operations in developing countries is inherently risky. Executing cost-effective UHP projects under such circumstances requires a higher degree of technological experience and management sophistication than executing similar operations in developed countries. QA/QC novices who function as virtual automatons, appraising all work with black and white, on-off switching logic, are generally unsuited for the challenge of managing UHP operations in developing countries.

Certainly, the lower salaries that novices receive allow QA contractors to submit competitive bids, but inevitably, you get what you pay for. Maximized value is never captured using QA representatives that are cheap on the front end but unequipped to provide quick-response solutions to the inevitable problems that surface in developing regions. Fast-track projects are unforgiving, and those who function as little more than go/no-go gauges are not a bargain in the long run.

Expat construction managers should have pipefitter/welding/controls backgrounds that provide actual production experience. Only by having production experience will construction managers be able to formulate practical solutions. Likewise, only by having a QA and analytical testing background can construction managers have the process experience to competently suggest solutions that will not jeopardize silicon processing.

The ongoing shift in QA organizational strategy will not only enable QA companies to survive, but will also allow clients to maximize the value that experienced QA professionals, serving as field construction managers, can bring to challenging UHP projects. If authorities with both QA and production expertise are deployed in construction management positions rather than as field monitors who are obliged to adhere strictly to the specs, flexible installation solutions with a latitude that serves both interests can be worked out. This new management structure will buffer the adversarial relationship between QA authorities and contractors, and expedite projects so that system integrity at the points of use can be preserved while production schedules move ahead with all possible speed.

Revamping the role of QA/QC will not eliminate all tensions. The quality imperatives inherent in UHP processes will continue to balance the drive toward uncontrolled, runaway installations and the physical demands of chipmaking. Commercial pressure and the limitations of developing regions require that future projects adopt more-efficient organizational structures in order to survive, much less thrive.