Introduction
The middle of line (MOL) represents a critical integration module in modern semiconductor manufacturing, bridging the active transistor devices formed during the front end of line (FEOL) and the macroscopic wiring networks created in the back end of line. The primary function of the MOL is to establish highly localized, low-resistance electrical connections—often referred to as local interconnects or contacts—to the source, drain, and gate terminals of the transistor. As device dimensions scale continuously, the geometric area available for these contacts shrinks dramatically, making MOL contact resistance a primary parasitic factor limiting transistor drive current and circuit switching speed. Historically, contact resistance was negligible compared to channel resistance, but in advanced nanoscale regimes, the physics of interfaces, carrier scattering, and thin-film material transport have placed MOL engineering at the core of process technology development.
Process map
See how a process flow is organized
Choose a technology node to explore its process map, module structure, and available steps. This opens the flow directory.
Physics & Mechanism
The fundamental physics of the MOL is governed by electronic transport across metal-semiconductor and metal-metal interfaces. At the interface between the contact metal and the heavily doped silicon or silicon-germanium source/drain regions, a Schottky barrier naturally forms. The specific contact resistivity of this interface depends critically on the active carrier concentration in the semiconductor; achieving ultralow contact resistivity requires driving active carrier concentrations near physical limits, shifting the dominant carrier transport mechanism from thermionic emission to quantum mechanical tunneling. Doping introduces donor or acceptor impurities that alter the Fermi level, but thermodynamic constraints and solid solubility limits constrain the maximum achievable active dopant concentration without inducing defect clustering or strain.
Within the metal plug itself, nanoscale thin-film electrical transport dictates conduction. When physical dimensions approach or fall below the electron mean free path of the metal, resistivity increases non-linearly due to electron scattering at surfaces and grain boundaries. This behavior is commonly modeled using the Mayadas-Shatzkes resistivity model, which quantifies the impact of grain boundary reflection and external boundary scattering. Furthermore, metal phase kinetics play a decisive role; for instance, tungsten (W) can form a high-resistivity metastable beta phase in ultrathin layers, whereas the low-resistivity alpha phase requires specific deposition conditions and energetic activation.
Process Principles
The fabrication of MOL structures relies on a controlled sequence of deposition, etching, and planarization steps designed to engineer atomic-scale interfaces. Contact formation typically begins with a surface pre-clean to remove native oxide, followed by deposition of a liner layer and diffusion barrier stack. Titanium (Ti) is commonly used as a liner to scavenge residual oxygen and form a low-resistance silicide at the source/drain interface, while titanium nitride (TiN) serves as a diffusion barrier. Due to the high aspect ratios of modern contact trenches and vias, physical vapor deposition is insufficient, necessitating conformal atomic layer deposition (ALD) or chemical vapor deposition (CVD).
In conventional W-based contacts, a thin ALD nucleation layer is deposited first to provide a uniform growth surface and prevent chemical attack on underlying materials during subsequent bulk CVD W fill. To minimize interface resistance between the MOL contact plug and subsequent BEOL interconnects, advanced integration schemes selectively remove barrier layers at the via bottom using directional reactive ion etching. Additionally, localized thermal activation techniques, such as laser annealing, provide brief, targeted heating to activate dopants at metal-semiconductor interfaces while preserving the overall device thermal budget.
Challenges & Failure Modes
The principal challenge in MOL integration is the physical scaling limit of barrier and nucleation layers. As contact critical dimensions shrink, constant-thickness Ti/TiN barrier and ALD nucleation films consume a larger fraction of the contact volume, reducing the cross-sectional area available for the high-conductivity bulk metal fill. Because barrier materials have higher bulk resistivity, the effective resistance of the scaled contact plug increases rapidly.
Electromigration and stress-induced voiding present severe reliability constraints. Under high current densities, electromigration damage can occur, usually in the form of void formation and hillock growth at the two ends, with a significant transport of metal atoms from one end of the line to the other . In scaled contact structures, high current densities aggravate this momentum transfer process. If interface cleaning prior to metal deposition is incomplete, or if precursor byproducts such as fluorine penetrate the barrier stack, interfacial oxidation and high-resistance contact failures occur. Furthermore, inadequate planarization of multi-material MOL stacks can create residual metal stringers, leading to inter-contact short circuits.
Technology Node Evolution
The evolution of MOL integration reflects the continuous drive to suppress parasitic resistance while scaling geometric features. In planar technology nodes, such as 28nm Planar Middle-of-Line Integration, a standard CVD W plug combined with a Ti/TiN liner stack provided sufficient process margin and low contact resistance. However, the introduction of 3D FinFET architectures at 14nm FinFET and 7nm FinFET nodes altered the contact topography, driving the adoption of ultrathin ALD barriers and modified contact geometry to maximize fill volume.
At advanced nodes, the volume penalty of the W nucleation layer became prohibitive, prompting a switch to cobalt (Co) for local interconnects and contact plugs. Cobalt possesses a shorter electron mean free path in bulk form, which reduces the relative resistivity increase caused by finite-size scattering effects when scaled to single-nanometer dimensions. Additionally, Co can be deposited without a thick resistive nucleation layer and undergoes thermal grain growth during post-deposition annealing, enabling lower line resistance and void-free reflow in high-aspect-ratio structures.
Related Processes
MOL module execution depends on upstream FEOL preparation and downstream planarization. Ion implantation establishes the ultra-high surface doping necessary for Schottky barrier narrowing at source/drain contacts. Thermal annealing steps repair lattice damage and transition dopants into active substitutional sites.
Following metal deposition, chemical mechanical planarization (CMP) is applied to remove overburden metal and barrier layers, isolating individual contact plugs and gate contacts. MOL CMP must polish heterogeneous material systems consisting of metals, nitrides, and dielectric oxides, requiring tailored slurry chemistry and mechanical parameters to prevent dishing or erosion.
Future Outlook
For future transistor architectures such as Gate-All-Around (GAA) nanosheets and complementary FETs (CFETs), MOL integration will undergo further material and structural transformations. Research on alternative metals demonstrates that ruthenium lines at extremely scaled linewidths show excellent electromigration behavior on single-damascene test vehicles . Molybdenum (Mo) and ruthenium (Ru) are being intensely investigated as barrierless or thin-barrier metal options due to their favorable bulk transport properties and oxidation resistance at reduced dimensions.
Area-selective ALD and bottom-up metallization strategies are also being developed to eliminate void formation and minimize volumetric barrier penalties. In parallel, monolithic 3D integration requires strict thermal budget management, accelerating the adoption of transient millisecond and nanosecond annealing technologies to optimize contact interfaces without degrading underlying active tiers.
References
State of the Art and Future Perspectives in Advanced CMOS Technology
H. Radamson, Huilong Zhu, Zhenhua Wu, Xiaobin He, Hongxiao Lin, Jinbiao Liu et al. · Nanomaterials
Silicon VLSI Technology - Full
James D. Plummer, Michael D. Deal, Peter B. Griffin
Silicon VLSI Technology · ISBN 978-0130850379