Random patterning variations impact semiconductor yield, costing billions

Fractilia, LLC, a leader in stochastics metrology, has revealed that uncontrolled random patterning variations at the latest technology nodes are costing semiconductor manufacturers hundreds of millions of dollars per fab in lost yield and delayed production. These devastating variations, known as stochastics, now represent the single most significant barrier to achieving expected yields in high-volume semiconductor manufacturing (HVM) at advanced nodes. Analysis by Fractilia and detailed in a white paper provides a roadmap for solving this problem by combining accurate measurement, probability-based process control, and stochastic-aware design strategies to recapture billions in unrealised value.

A gap has emerged between the size of features that can be patterned in research and development and the size of features that can be reliably manufactured at high volumes at historical expected yields. This resolution gap is largely caused by stochastic variability, which is the random variability in the behaviour of molecules, light sources, and even atoms in the materials and equipment used in semiconductor lithography. Unlike other forms of process variability, stochastic variability is inherent to the materials and technologies used in the manufacturing process and must be addressed using a probability-based analysis that differs from current process control approaches.

“Stochastic variability is contributing to multibillion-dollar delays in introducing advanced process technology into high volume manufacturing,” stated Chris Mack, Chief Technology Officer of Fractilia. “Yet conventional process control approaches have been ineffective in addressing these random effects. Closing the stochastics gap requires completely different methodologies that device makers need to validate and adopt to move advanced process technologies into high volume manufacturing successfully.”

According to Mack, the most significant limitation to growth in the electronics industry is stochastic variability. “We have seen our customers make dense features as small as 12 nanometers in research and development, but when they try to move it into manufacturing, stochastic failures are affecting their ability to achieve acceptable yield, performance, and reliability,” said Mack.

Historically, stochastic variability has not had a significant impact on high-volume manufacturing yield. Stochastic effects were small enough compared to critical feature sizes that the probability of yield-killing stochastic defects was low. However, with the substantial increase in lithography capability enabled by the adoption of EUV and high-NA EUV technologies, stochastic variability consumes a much larger percentage of the error budget in advanced manufacturing processes.

Fortunately, this stochastic gap is not a fixed size. Fractilia details the causes of this stochastic gap and provides solutions for addressing it, including stochastic-aware device design, material advancements, and stochastic-aware process control. “The stochastics gap is an industry-wide problem,” stated Mack. “This issue can be minimised and controlled, but it all starts with accurate stochastic measurement technology.”