Perspectives
03/01/2003
SST editors ask industry execs about advanced process control
Recently, Solid State Technology asked several key industry executives: How do you perceive advanced process control (APC) developing (and being funded) in wafer fabs?
 Dick Deininger, director of manufacturing technology, AMD |
At AMD, where APC is already a key element in manufacturing, APC implies fully automatic, model-based process control, with both single-loop (within tool) and cascaded controllers (between tool). APC also uses the latest feedback and feed forward control technology. Fault detection and classification (FDC) is a component of APC. FDC applications are used to monitor the real-time performance of critical process steps. AMD's FDC solution assesses the current tool state versus a known good state on a run-by-run basis. This allows us to quickly identify and correct abnormal processing. The result is lower product jeopardy and faster cycle time.
 Thomas Sonderman, director of APC technology, AMD |
It is our belief that the majority of the industry is currently using tool-based statistical process control together with automatic tool shutdown — a key perquisite for APC. Over time, we believe that there will be a significant migration toward more APC applications. The demand for this capability will increase as smaller feature sizes and technology requirements demand greater precision from the integration and control of individual process tools.
OEMs must continue to provide increasingly robust data processing and to deliver tools with APC-enabled functionality. IC manufacturers will be required to implement fab-wide control systems that address both inter- and intra-tool performance objectives. This requires a highly functional MES, dynamic recipe management systems, leading-edge engineering analysis systems, and a fab-wide APC framework.
The implementation of APC also requires a major shift from traditional fab management practices. Management must feel comfortable with the use of automation technology and the concept of controlling a process to an objective function (as opposed to a static recipe). More important, management must be willing to understand the dynamic nature of leading-edge IC processing and the need for fully automated control technology. This understanding will convince it to make the necessary infrastructure investments required to take advantage of fab-wide control.
At the 90nm and below technology nodes, fab-wide APC will be a fundamental requirement toward ensuring flexible manufacturing with the highest levels of fab productivity and the greatest yields.
 Lee Smith, senior VP of business development, Therma-Wave Inc. |
The revenue that was lost due to process excursions over the past 30 years is staggering, perhaps several billion dollars/year for the IC industry as a whole. A beneficial consequence of APC should be a major reduction in excursions, and that alone will be a significant boost to the productivity of fabs. Moreover, the benefit of APC to 300mm fabs is 2.5¥ that for 200mm fabs. Another factor is that with 300mm equipment comes the enabling use of standard equipment frontend design, allowing attachment of integrated metrology (IM) to process tools without custom engineering.
Looking forward, the macro trend of IC feature density increasing at a relentless pace, and now pushing deep into the sub-wavelength regime, means that more vigilance must be supplied by metrology to achieve and maintain leading-edge chip performance. As a specific example, more etch and CMP processes are making "blind" stops, relying on metrology for control. For another example, gate trim processing is making complex 3-D adjustments to device structures, relying on optical CD for control. The role of metrology is increasing, no doubt about it.
However, the business model for APC is one of the first-order issues that remain to be worked out. This three-way interaction involves two large players — the process toolmaker and the IC maker — and the usually smaller metrology supplier. If leverage is successfully applied to the metrology supplier such that it is unable to make respectable gross margin on an IM system, the end result will be an unsustainable relationship. Unfortunately, IM suppliers may be so competitive that they allow such a subsistence-diet model to get started.
Alternatively, if IM suppliers sell directly to IC makers to eliminate the "margin stack-up" problem, there can be major issues of gaining and sustaining cooperation from the process tool maker, and of managing the questions of which party is responsible for various operational issues that impact overall success. The business and control issues are arguably the most complex and challenging hurdles yet to be overcome for APC.
An event that perhaps accelerated APC and IM into the mainstream was the recent process tool suppliers' strategy to sell complete process modules, such as the patterning modules encompassing track and etch — with yield and performance guaranteed by the process tool supplier. This moved the responsibility for process success from the IC maker to the process tool supplier.
Another question often discussed is "can the performance of IM equal that of a stand-alone, traditional metrology system?" If IM technology is compared to the same, single technology of a stand-alone system, it is plausible that IM performance can, over time, approach that of the stand-alone tool. Even if IM has only 50% of the repeatability of the stand-alone system, however, it can still perform the very rewarding function of giving virtually immediate warning of process drift or a crash and thus prevent excursions. Understanding the value components of APC and IM is an issue that will settle itself out over the next few years.
 James L. Brissenden, president and CEO, INFICON |
The adoption of APC in the semiconductor industry is driven by geometrically increasing process complexity and cost of failure. Underlying these are tighter process parameters, fundamentally new materials, and larger, more expensive wafers. These factors combined with signs that the semiconductor industry is becoming more mature, characterized by product commoditization and price pressure, are creating a new urgency to address manufacturing efficiency and equipment use.
APC is most successfully implemented when the first priority targets are mission-critical and problem-prone processes. This Pareto analysis-driven prioritization maximizes yields and equipment uptime in the most cost-intensive fab processes. APC solutions for these processes are dependent on in-depth understanding of process chemistries and physics. This implies application-specific expertise and real-time process monitoring, analysis, and control. Initially, these APC initiatives involve optimizing individual processes, but they are rapidly evolving into feedback and feed-forward control loops for optimizing related blocks of processes.
APC can only be funded from the cost savings it produces. Thus, APC investments need to be targeted at mission-critical and "bottleneck" processes where increased control can create large and rapid paybacks.
Among the potential benefits provided by these APC initiatives, making more product with fewer tools is probably the most important. The cost of equipment to produce new generations of chips is at an unsustainably high level given current and probable future IC prices. Historically, semiconductor equipment capital expenditures have represented about 25% of IC revenue. The economics of semiconductor manufacturing today dictate that equipment spending as a percent of revenue be significantly reduced to provide IC manufacturers adequate profitability. The obvious solution to this problem is to use APC to raise semiconductor equipment utilization from today's 60–90%-plus, which is representative of most mature process industries.
Who is driving the adoption of APC? Large IDMs and foundries have the biggest financial incentive and are currently most active. However, it is unlikely that the majority of IC manufacturers have the internal resources and know-how for a total and effective APC implementation. This will translate into more pressure exerted on the semiconductor OEMs to deliver APC-enabled process equipment and more outsourcing to smaller software and sensor suppliers that provide unique process-oriented solutions.
 John Heaton, president and CEO, Nanometrics Inc. |
I see three ways APC will be funded in future fabs. First, top-tier IDMs and foundries will internally develop their own code and implement it through factory automation software already in place. Second, major equipment companies with diversified product lines will begin packaging APC, integrated metrology, and other processing tools in order to offer customers a complete process module, such as a patterning or copper module. Third, it will become increasingly common for small companies to offer software packages to control specific machines that identify immediate payback. In all three cases, the customer will absorb the costs, but development money will generally be funded by the equipment community.