2017: A transformative year for through-silicon vias (TSVs)?
By Frédéric Raynal, aveni
What do we predict will be major technology drivers for 2017? While bleeding-edge foundries and integrated device manufacturers (IDMs) are publicly discussing the challenges to manufacture devices at the 10nm and 7nm nodes, we know there is much they are not talking about. Such as, how long will copper remain viable for metallization? When will radical material and the associated structural changes require implementation? Will 2017 be the transformative year when through-silicon vias (TSVs) reach mainstream adoption?
We believe logic devices as we know them today will continue to be manufactured, but with some caveats. FinFETs (fin field-effect transistors) still will be used in front end of line (FEOL). Likewise, copper metallization for back end of line (BEOL) is extendible to the 3nm node, provided dimensions do not physically limit the material choice. Conventional acidic copper electrochemical deposition (ECD) is reaching its limits at the 10nm and 7nm nodes, with high likelihood that only alkaline chemistries will meet requirements.
Cobalt (Co) electroplating is in active development to replace tungsten chemical vapor deposition (CVD-W) for middle of line (MOL) ≤7nm. Device designers are driving this material change because Co can be electrodeposited with a bottom-up approach, whereas CVD-W’s inherently conformal deposition becomes a liability at smaller geometries.
On a related note, some device manufacturers are moving to Co for BEOL at ≤10nm, which imposes requisite cost increases, delays due to development and reduced chip performance. Concerns about copper electromigration are driving this change. Marginal seed layers have compromised the reliability of devices metallized with acidic copper chemistries for ECD. Alkaline chemistries can delay transition from copper to unconventional metallization schemes, while maintaining high device reliability, to at least the 3nm node. However, 2nm is widely considered to be the technology disruption point for new integration schemes, be they carbon nanotubes or others.
Finally, we won’t predict that 2017 will be the breakthrough year for TSVs. They will remain as niche applications: most notably, high-bandwidth memory and CMOS image sensors.