Making Interconnects Faster

How cobalt compares to copper, and why it matters.

popularity

In integrated circuits, interconnect resistance is a combination of wire and via resistance. Wire resistance of a conductor depends on several factors, one of which is the electron scattering at various surfaces and grain boundaries. Via resistance, on the other hand, is a function of the thickness or resistivity of the layers at the bottom of the via through which current must travel.

This comparative animation of 15nm critical dimension (CD) dual damascene structures illustrates how electrons traveling through copper (Cu) encounter numerous grain boundaries and surfaces that cause scattering and higher resistance. As CDs scale below the mean free path of Cu (39nm), electron scattering increases and this causes Cu wire resistance to rise. In addition, the Cu interconnect requires a liner as well as a barrier layer, reducing the volume of Cu and resulting in higher via resistance.

By comparison, the lower mean free path of cobalt (Co) causes less electron scattering, leading to better resistance scaling as CDs approach 10nm to 12nm. Further, only a single, 1nm-thick barrier film is required for Co interconnects, which minimizes via resistance.

Co’s lower resistance enables signals to travel faster and for a longer distance. In the race car analogy within the animation, this is represented by the Cu car completing three circuits, whereas the Co car completes four circuits in the same time, thus traveling faster and farther with a given amount of power.



3 comments

Jim Lloyd says:

Not sure the smaller mean free path contributes to better conductivity, that doesn’t really make sense. The mean free path is determined by the scattering. I think where Co brings an advantage is that the Fuchs scattering is less due to more specular surface/interface reflections.

Qiang Huang says:

My understanding is the smaller mean free path contributes to “having less increase of the overall resistance when the structure dimension shrinks”. It does not make the conductivity better, but makes it less worse.

Jim Lloyd says:

But that would indicate that the beginning conductivity would be worse and there would be no advantage to use Co. Unless the main contributor there is the lack of a liner.

Leave a Reply


(Note: This name will be displayed publicly)