Molybdenum Process

The molybdenum process deposits Mo as an emerging replacement for tungsten in 3D NAND word lines, and the problem driving that swap is a different flavor of resistance crisis than the one covered in the ruthenium process. Ruthenium's story is about wires getting too narrow; molybdenum's is about word lines getting too long — stacking 3D NAND another layer taller doesn't make any single word line narrower, it just adds one more full-length conductor to a stack where every layer already has to snake across the entire array.

Sacrificial SiN stack etched
100+ alternating oxide/nitride layers
Narrow slit etch
Vertical access channel through the whole stack
Replacement gate: pull nitride, flow in metallateral fill through the slit, layer by layer
Wet-etch nitride out
Hot phosphoric acid, through the slit
Mo (or W) flowed in laterally
Fills every word-line cavity from the slit inward
Lower bulk resistivity
Mo beats W across very long word lines
Reduced RC delay
Across the full word-line length, not a narrow pinch point

3D NAND's word-line resistance problem scales with layer count and array size, not with linewidth, which is exactly the opposite emphasis from ruthenium's narrow-line crossover story. Every additional layer in a 3D NAND stack adds another full-length word line running the entire width of the memory array, and every one of those word lines has to be driven fast enough to read and program the memory cells sitting along it. As stacks climbed past 100 and then 200+ layers, the sheer cumulative resistance and capacitance of that many long, thin conductors — not any single conductor shrinking below a mean free path — became the dominant source of word-line RC delay. Tungsten's bulk resistivity, perfectly adequate for the shorter interconnects and contacts covered in the tungsten process, started limiting how fast a word line that long could actually switch.

3D NAND Replacement Gate: Word Lines Flowed in Through a Slit every layer gets its own full-length word line — resistance scales with stack height and array width, not linewidth narrow slit (etch + fill access) ⋮ repeated 100+ times up the full stack word-line metal — full array width another full-length word line metal flows in laterally from the slit Mo's lower bulk resistivity helps every one of these full-length conductors at once — not just the narrowest ones

Molybdenum's lower bulk resistivity than tungsten directly reduces that word-line delay without needing any size-effect crossover argument at all — this is a straightforward bulk material swap, not a nanoscale transport-physics story. Where ruthenium only wins once a wire shrinks below its own short mean free path, molybdenum's resistivity advantage over tungsten applies across ordinary word-line dimensions, because the comparison here is simply which metal conducts better in bulk, at a scale where neither metal is deep into its own size-effect regime. That makes molybdenum's adoption path more straightforward in one sense — no crossover width to hit — but it still has to clear the same replacement-gate process compatibility bar tungsten already cleared: flowing cleanly and completely through the narrow slit into every one of a hundred-plus thin lateral cavities without voiding, the 3D-NAND-specific version of the gap-fill reliability problem the tungsten and copper processes cover for contacts and vias.

That lateral, slit-fed fill geometry is genuinely unlike anything in planar BEOL, and it's a large part of why the switch from tungsten to molybdenum isn't just "swap the gas and go." A contact or via fills vertically, straight down from an opening directly above it; a 3D NAND word line fills horizontally, flowing sideways from a narrow vertical slit into a cavity that can run for a significant lateral distance before reaching the far side of the array — a much longer, more void-prone fill path than any contact plug has to manage. Molybdenum's deposition chemistry has to achieve the same void-free, complete fill tungsten's WF6 process already proved out for that lateral geometry, which is as much the engineering challenge in adopting it as the resistivity improvement itself.

Read the molybdenum process as the word-line counterpart to ruthenium's logic-interconnect story: both metals are being substituted into advanced-node flows specifically because copper and tungsten's own properties stopped being good enough at the new scale each technology reached — but ruthenium's crossover is about wires getting narrower than an electron's mean free path, while molybdenum's is about word lines simply getting longer, one stacked 3D NAND layer at a time, with total resistance climbing in direct proportion to how tall the stack has grown.

Take molybdenum process further

Ask the copilot about this term, or have our engineers assess it against your process.