By David W. Price and Douglas G. Sutherland
Introduction to the Series:
This is the first in a series of 10 installments which will discuss fundamental truths about process control—inspection and metrology—for the semiconductor industry. By fundamental, we imply the following:
- Unassailable: They are self-evident, can be proven from first principles, or are supported by the dominant behavior at fabs worldwide
- Unchanging: These concepts are equally true today for 28nm as they were 15 years ago for 0.25µm, and are expected to hold true in the future
- Universal: They are not unique to a specific segment of process control; rather they apply to process control as a group, as well as to each individual component of process control within the fab
Each article in this series will introduce one of the 10 fundamental truths and discuss interesting applications of these truths to semiconductor IC fabs. Given the increasing complexity of advanced devices and process integration, process control is growing in importance. By understanding the fundamental nature of process control, fabs can better implement strategies to identify critical defects, find excursions, and reduce sources of variation.
The first fundamental truth of process control for the semiconductor IC industry is:
You can’t fix what you can’t find. You can’t control what you can’t measure.
While it’s true that inspection and metrology systems are not used to make IC devices—they do not add or remove materials or create patterns—they are critical for making high-yielding, reliable devices. By finding defects and measuring critical parameters, inspection and metrology systems monitor the hundreds of steps required to manufacture a device, ensuring the processes meet strict manufacturing specifications and helping fab engineers identify and troubleshoot process issues when there is an excursion. Without inspection and metrology, it would be near impossible for fabs to pinpoint process issues that affect yield. However, it’s not enough to simply “find” and “measure”—a fab’s process control strategy needs to be capable and cost-effective.
Capable inspection and metrology strategies find and measure the defects and parameters that affect device yield. Cost-effective inspection and metrology is performed at the lowest total cost to the factory, where total cost is the sum of the cost of lost yield plus the cost of process control.
First, Make it Capable:
If you can’t find it, you can’t fix it. At the heart of this truth is the understanding that, above all else, a fab’s inspection and metrology strategy must be capable. It must highlight the problems that are limiting baseline yield. It must also provide actionable information that can enable fabs to quickly find and fix excursions.
Figure 1. You can’t fix what you can’t find. And you can’t control what you can’t measure. Left: P-MOS SiGe critical dimension measurement. Right: Fin patterning particle leading to a Fin Spire defect at post dummy gate etch. Source: KLA-Tencor
We emphasize this need for capability first because we have observed that some fabs are too quick to sacrifice capability for cost reductions. No strategy is cost-effective if it doesn’t accomplish its fundamental objective.
Below are specific questions that can help fab management evaluate the capability of its process control strategy:
- Are you finding all sources of your defect-limited yield? Are you finding these in-line or at end-of-line?
- Does your defect Pareto have sufficient resolution of the top yield-limiters in each module to direct the most appropriate use of factory engineering resources?
- Have you fully characterized all of the important measurements and defect types (size range, kill ratio, root cause, solution)?
- Do you understand the most probable excursion scenarios? What is the smallest excursion that you absolutely must detect at this step? How many lots are you willing to have exposed to this excursion before it is detected?
- Are you inspecting and measuring at all the right steps? Can you quickly isolate the point of formation for excursions? Can you quickly disposition potentially affected lots?
- Does a particular defect signature become confused by defects added at subsequent process steps? Or do you need separate inspections at each step in order to partition the problem?
- Do you have overlapping inspections to guard against the high-frequency, high-impact excursions?
- What is the alpha risk and beta risk for each inspection or measurement? How are these related to the capture rate, accuracy, precision, matching and more?
Then, Make it Cost-Effective:
Once a capable strategy is in place, then a fab can start the process of making it cost-effective. The best known method for optimizing total cost is usually adjusting the overall lot sampling rate. This is generally preferred because the capability remains constant. In some cases, it may be possible to migrate to a less sensitive inspection (lower cost of ownership tool or larger pixel size); however, this is a dangerous path because it re-introduces uncertainty (alpha/beta risk) that reduces a fab’s process control capability. This concept will be discussed in more detail in our next article on sampling strategies.
Finally, it is worth pointing out that it is not enough to implement a capable strategy. The fab must ensure that what was once a capable strategy, stays a capable strategy. A fab cannot measure with a broken inspection tool or trust a poorly maintained inspection tool. Therefore, most fabs have programs in place to maintain and monitor the ongoing performance of their inspection and metrology tools.
By optimizing process control strategy to be capable and cost-effective, fabs ultimately find what needs to be fixed and measure what should be controlled—driving higher yield and better profitability.
About the Authors:
Dr. David W. Price is a senior director, and Dr. Douglas Sutherland is a principal scientist at KLA-Tencor Corp. Over the last 10 years, Dr. Price and Dr. Sutherland have worked directly with more than 50 semiconductor IC manufacturers to help them optimize their overall inspection strategy to achieve the lowest total cost. This series of articles attempts to summarize some of the universal lessons they have observed through these engagements.
Check out other Process Watch articles: “Exploring the Dark Side,”“The Dangerous Disappearing Defect,” “Skewing the Defect Pareto,” “Bigger and Better Wafers,” “Taming the Overlay Beast,” “A Clean, Well-Lighted Reticle,” “Breaking Parametric Correlation,” “Cycle Time’s Paradoxical Relationship to Yield,” and “The Gleam of Well-Polished Sapphire.”
Thanks, great idea for this 10 part series “fundamental truths of process control” in our industry including a great definition of what you mean by “fundamental truths”
And behind door number one!!!!!
“You can’t fix what you can’t find. And you can’t control what you can’t measure”
I can’t wait for the other nine…well into 2016, the tipping point year (my opinion) for this radically as we move below 28nm in volume manufacturing (for bleeding edge products).
I started making IC’s in 1960 when government support jump-started this entire industry with Minuteman IC’s and beyond (the bipolar era). Some fundamental truths most likely worked even then. When we see all 10, I will try to see how far BACK those truths were recognized. Diode era? Transistor era? Most of those companies were not able to lead the IC era. In fact Diode leaders were not Transistor era leaders. That first one, “FT1”? seems universal for all manufacturing processes historically. But our industry certainly pushed process control to extremes of precision and robustness while doubling product complexity and reducing costs as the only “deflationary” industry.