Transistor Scaling Gated by Contact Resistance
At Applied Materials, we are working closely with our customers to help them solve the semiconductor design and manufacturing challenges of the AI Era.
For decades, the industry relied on classic 2D scaling to drive simultaneous improvements in power, performance and area/cost (PPAC), but that playbook has become more challenging.
The advent of EUV lithography has helped overcome some of the 2D scaling challenges, but at the same time it has brought to light other problems that are becoming more critical as transistor feature sizes shrink. In this blog, the first in a series, I will explore the issue of contact resistance and explain why it is becoming a growing impediment to transistor power and performance scaling—and why a breakthrough in materials engineering is needed.
The contacts are the first and one of the smallest levels of wiring that connects the transistor to the rest of the wiring in the chip. The contacts help drive electrons across the channel once the gate activates the transistor into its “on” state (see Figure 1).
For many years, transistor contacts have been formed by creating vias within a dielectric layer, cladding the vias with liner-barriers and nucleation layers and filling the remaining portion of the via with tungsten (W), the contact metal of choice due to its low resistivity.
The liner-barrier is typically composed of titanium nitride (TiN) which is good at adhering to the sidewall of the via and preventing defects like delamination and voids. Because tungsten doesn’t naturally grow on TiN, a W nucleation layer is deposited before the W bulk fill. The TiN liner and W nucleation layer are like a mold that defines the space for the W contact (see Figure 2). Although this mold takes up a large percentage of the via’s space, it does little to conduct electrons because its resistance is much higher than that of bulk W.
With the advent of EUV lithography, the contact vias can continue to scale. But contact power and performance are hitting the wall: the “mold” is now taking up more volume than the contact metal, and contact resistance is growing (see Figure 3). Without a breakthrough in contact metal deposition, there is no benefit to shrinking: the transistor is like a Formula 1 race car stuck in rush hour traffic.
In my next blog, I’ll describe some of the techniques Applied Materials has developed to relieve the contact resistance bottleneck—and why more work is needed to enable new ways to shrink.