Alternative Contact and Interconnect Metals represent the shift away from tungsten contacts and copper local wires at advanced CMOS nodes — adopting cobalt (Co), ruthenium (Ru), and molybdenum (Mo) to overcome the scaling limitations of traditional metals, where tungsten's high bulk resistivity and copper's large grain boundary and surface scattering at nanometer dimensions create unacceptable resistance increases that alternative metals can partially solve through thinner barriers, barrier-free integration, or favorable electron transport properties.

Why Traditional Metals Fail at Nanoscale

• Tungsten (W) Contacts: W has been the standard contact fill metal since the 0.5 μm node. But W requires a TiN/Ti adhesion/barrier layer (3-4nm) that占s an increasing fraction of the contact volume as contact diameter shrinks below 15nm. W itself has high bulk resistivity (5.3 μΩ·cm), and at nanoscale dimensions, the effective resistivity further increases. The combined barrier + fill resistance becomes a major performance limiter.

• Copper (Cu) Wires: Cu (1.7 μΩ·cm bulk) requires a Ta/TaN barrier (3-5nm) and Cu seed layer. At wire widths below 20nm, the barrier consumes 40-50% of the wire volume, and the remaining Cu suffers severe grain boundary and surface scattering (effective resistivity 5-8 μΩ·cm). Cu's advantage over alternative metals diminishes at sub-20nm dimensions.

Cobalt (Co)

Co (6.2 μΩ·cm bulk) has higher bulk resistivity than Cu but advantages at nanoscale:

• Thinner Barrier: Co can use a thin TiN liner (~1nm) or even direct deposition on dielectric in some integrations. More metal fill volume per given contact hole diameter.
• Better Fill: CVD Co provides superior void-free fill in high-aspect-ratio contacts compared to PVD + electroplated Cu or CVD W.
• First Adoption: Intel used Co for M0 and M1 (local interconnect) at the 10nm node (Intel 7). TSMC uses Co contacts at N5 and below.

Ruthenium (Ru)

Ru (7.1 μΩ·cm bulk) is the leading candidate for the tightest-pitch wires at N2/A14 and beyond:

• No Barrier Required: Ru does not diffuse into dielectrics and provides its own adhesion — no barrier or liner needed. 100% of the wire cross-section is conductive metal.
• Low Size Effect: Ru has a shorter electron mean free path than Cu (6.7nm vs. 39nm), meaning surface/grain boundary scattering increases its resistivity less at narrow dimensions. Below ~10nm width, Ru can have lower effective resistivity than Cu+barrier.
• Integration: ALD and CVD Ru processes are being qualified for selective and conformal deposition.

Molybdenum (Mo)

Mo (5.3 μΩ·cm bulk, same as W) has an extremely short electron mean free path (1.4nm), making it the most resistant to size-effect scattering. At sub-10nm wire width, Mo's effective resistivity stays close to its bulk value — potentially the best metal for the narrowest wires. Under evaluation at multiple foundries for M0-M2 at the 2nm node and beyond.

Alternative Interconnect Metals represent the recognition that the best bulk conductor is not always the best nanoscale conductor — that at the dimensions of advanced CMOS, the boundary conditions matter more than the bulk property, making metals with shorter electron mean free paths and thinner barriers the practical winners despite higher intrinsic resistivity.