Continuous scaling of transistors has forced the semiconductor industry to reinvent fundamental transistor architecture and manufacturing processes.
The most advanced transistor that is currently commercially available is FinFET—a type of multigate device for high speed/high-density processors.
Just like earlier generations of transistors, FinFET is hitting its limits in terms of scaling and leakage current, and it may need to be replaced with a new architecture in the near future.
This silicon die for a GPU hides tiny transistors made from FinFET technology.
Every time the major semiconductor manufacturers scale to a new technology node, they need to reinvent the manufacturing process, transistor architecture, or both. The current commercialized transistor node is 7 nm, with 5 nm set to begin at full scale production sometime in 2020, and both transistor nodes are based on FinFET technology. New products at the next technology node will have even higher density, and whether the transistor architecture will need to change again is an open question.
As transistors have continued to scale downwards, FinFETs have proven their worth in advanced microprocessors. The CMOS process currently dominates in PLDs and has helped ensure these devices use less power than they did in decades past. Advanced technology in development will continue to be computationally intensive, and CMOS FinFETs are likely to remain the primary transistor architecture for these advanced applications.
History of FinFET Technology
FinFET technology was introduced after a gradual transition from planar architecture to a vertically-oriented gate architecture. Although MOSFETs were originally introduced in 1960, a double-gate thin-film transistor was patented in 1980, where the channel in a FET is contacted on each side by a double-gate electrode structure. The double-gate structure later evolved into a nonplanar multi-gate structure and, ultimately, into the FinFET structure used in modern technology nodes. The first commercially available products to use FinFETs were manufactured by Intel at the 22 nm node.
A multi-gate structure continues building the gate electrode structure around the gate region, and the gate region is extended out of the silicon to form a nonplanar transistor structure. FinFET technology simply completes the structure by enclosing the entire gate region with a wrap-around gate electrode. FinFET is often referred to as a “trigate”, as it surrounds the channel on 3 sides (out of 4 possible). There is some debate as to whether FinFETs and trigates should be considered the same device. The structure of a modern FinFET is shown below.
Structure of a FinFET transistor.
Today, modern microprocessors, GPUs, and other high-density processors use FinFET technology as the fundamental transistor architecture. In this structure, the gate electrode can be made from polysilicon or a metal alloy, and there is a thin insulator between the channel region and the gate electrode. This was traditionally an oxide layer (e.g., in MOSFETs), although high-k dielectrics are now being used to combat short-channel effects. FinFETs can also be formed with multiple fins placed in parallel, which gives larger gain with similar control over short-channel effects.
This need to combat short-channel effects—where excessive leakage current passes between the source and drain regions when the channel size is small—underlies the motivation for using FinFETs in modern microprocessors.
SOI vs. Bulk FinFET Technology
Like any other area of electronics, there are different varieties of FinFETs, each with their own advantages. In the structure shown above, the source, gate, and drain can sit on an insulating layer (silicon on insulator, or SOI) or on bare silicon (bulk). SOI provides self-alignment during growth and strong isolation between neighboring FinFET structures. Both SOI and bulk FinFETs can use metal or polysilicon gates.
Why Use FinFETs?
There are many reasons the industry has transitioned from 2D planar transistor architecture to other styles of transistors (including 3D FinFET transistors)—all of which center around controlling leakage current. As transistors have scaled smaller, electrons have a higher probability of passing between the source and drain regions due to quantum tunneling. This leads to higher leakage current when a source-drain voltage is applied, even if the gate is turned off.
The wrap-around structure of a FinFET gives designers more control over leakage current in the following ways:
Stronger depletion: By wrapping the gate electrode around the channel, short-channel effects are reduced as the gate is driven more fully into different operation modes. This occurs because the electric field in the channel can be made more uniform through the wrap-around gate structure. In addition, there is no doping variation throughout the body, which reduces threshold voltage variations due to substrate bias (body effect).
Adaptable to existing processing steps: FinFET is not the same as CMOS, as it is a non-planar architecture, but the same process steps can be used for fabrication. The main challenge has focused on EUV lithography, rather than on reengineering existing processing steps.
New scaling parameters: FinFET technology is allowing further scaling beyond planar architecture by introducing the fin thickness, fin height, and gate length as new scaling parameters. Leakage current is better suppressed if the fin thickness is less than the gate length.
In addition to these basic advantages, the geometry of a FinFET can be used to tune short channel effects (e.g., triangular or curved fins). This leads to future variations on FinFET technology, where the channel region is lifted off the substrate altogether.
Beyond FinFET and Below 5 nm
The natural next step to continue scaling transistors, especially below 5 nm, is to continue wrapping the gate around the fourth side of the channel region. This structure is called gate-all-around FET (GAAFET). Some possibilities include growing nanosheets or nanowires from silicon or III-V materials and using the structure as the channel region. Other possibilities include using alternative materials (e.g., graphene) in the standard FinFET structure. No matter what architecture is used, designers will need the right tools to design, model, and integrate their FinFET circuits into larger systems.
Over time, the progression of transistor architecture has gone from planar to transistors to FinFET technology. To continue taking control of short-channel effects, GAAFET structures may aid continued scaling.
When you’re ready to start designing circuits with FinFET technology, you need a complete set of circuit design, simulation, and analysis tools. Curate a complete circuit design application for creating cutting-edge technology, and define transistor models as circuit blocks to begin creating your new system.
If you’d like to keep up-to-date with our System Analysis content, sign-up for our newsletter curating resources on current trends and innovations. If you’re looking to learn more about how Cadence has the solution for you, talk to us and our team of experts.