Tungsten deposition fills M0 contact openings across managed barrier and nucleation interfaces, maximizing conductive volume for reliable electrical connection .
After the planarization of tungsten and Ti/TiN contact plugs via CMP, the Metal 0 (M0) W Deposition step establishes the first local in
terconnect layer in the nanoscale BEOL flow . M0 serves as the critical bridge linking high-aspect-ratio contact plugs to the subsequent global routing layers . Unlike the Metal 1 through Metal 4 steps in this flow, which utilize copper electroplating within a dual-damascene architecture , this M0 step employs blanket tungsten deposition followed by subtractive patterning . Tungsten is selected for this intermediate local routing because it possesses excellent thermal stability, the highest melting point among metals, and perfect resistance to electromigration . Furthermore, implementing a subtractive W etch for M0 circumvents the severe integration challenges associated with fine-pitch copper CMP at the device level . The deposition process fundamentally relies on the chemical reduction of tungsten hexafluoride (WF6) to form solid metallic tungsten . Because bulk chemical vapor deposition (CVD) of W struggles to nucleate directly on typical dielectric or barrier surfaces, an ultra-thin atomic layer deposition (ALD) nucleation layer is first grown using alternating pulses of WF6 and a reducing agent such as silane (SiH4) or diborane (B2H6) [P1, P4]. This nucleation layer initiates uniform island growth, allowing the subsequent CVD phase—driven by the hydrogen (H2) reduction of WF6—to proceed rapidly and uniformly . The thermodynamic conditions and precursor choices during these steps strictly dictate the resulting crystalline phase of the film . The microstructural evolution must be carefully controlled to favor the stable α-phase (body-centered cubic) over the metastable β-phase (A15 structure), because the α-phase exhibits significantly lower intrinsic resistivity [P1, P4]. The integration of W at the M0 level necessitates stringent management of the barrier and nucleation interfaces to optimize the effective conductive volume . Conventionally, a titanium nitride (TiN) adhesion layer is required to prevent the aggressive WF6 precursor from attacking underlying oxide layers or any exposed titanium, which would otherwise generate volatile TiF4 and lead to massive structural defects . However, because TiN and the nano-crystalline ALD W nucleation layer possess inherently high electrical resistivities (700-800 µΩ·cm and 100-150 µΩ·cm, respectively), their physical thicknesses must be minimized to maximize the volume available for the low-resistance bulk α-W . Modulating the deposition temperature and the specific precursor gas ratios directly influences the nucleation density and the resulting surface roughness . Consequently, optimizing these kinetic parameters ensures that the blanket W film remains smooth enough for the subsequent high-resolution photolithography and subtractive etching steps (Engineering Practice). For a nanoscale Back-Illuminated (BSI) CMOS Image Sensor, minimizing the overall stack resistance is essential for maintaining high-speed signal readout and low dynamic power consumption . As device dimensions scale, the M0 interconnect resistance becomes a dominant contributor to the total parasitic series resistance due to enhanced electron scattering at interfaces and grain boundaries . Therefore, achieving an optimal W grain structure during this deposition step directly impacts the RC delay of the local interconnect network, serving as a critical determinant of sensor performance .
Sign in to continue through all 417 steps