Node Separation

Node separation, also called bottom electrode separation or BE etch-back, cuts the TiN bottom electrode layer between neighboring capacitor cylinders after the mold dip-out has already left them standing free, because the metal that lined every storage node hole was deposited as one continuous film across the whole wafer and has to be physically divided before each cylinder can function as its own electrically independent capacitor. A deposited electrode film doesn't come pre-separated into individual cells — it blankets the entire array, bridging cylinder to cylinder along the surface between them, and until that bridging material is removed, every capacitor on the wafer is still electrically connected to its neighbors rather than storing its own isolated charge.

Choosing a dry etch-back versus a CMP-based separation comes down to how the free-standing cylinders from the mold dip-out can tolerate being handled, since this step happens on a structure with nothing but thin nitride support bridges holding it upright. A dry plasma etch-back removes the connecting TiN with directional control and no mechanical contact, which avoids any risk of physically disturbing the delicate, unsupported cylinder array; a CMP-based approach instead physically polishes the connecting material away, which can achieve excellent planarity but introduces direct mechanical contact with a structure that, at this exact point in the flow, is about as fragile as anything in the entire DRAM process gets — the choice between the two is as much about what the fragile array can survive as it is about which method separates more cleanly.

Any residual bridging film left behind by an incomplete separation doesn't fail quietly — it shows up directly as electrical crosstalk between cells that should be fully isolated. A storage cell that leaks charge into its neighbor through an incompletely separated electrode behaves exactly like a cell with a retention problem, which can send a yield investigation chasing a dielectric-quality or leakage-mechanism explanation when the actual root cause is this etch not having cleared completely — this is part of why node separation gets inspected with the same statistical rigor as the hole etch and dip-out steps that precede it, even though it looks, on the process flow diagram, like a comparatively simple cleanup pass.

Because this step runs immediately after the most structurally delicate point in the entire flow, it's qualified as tightly coupled to the mold dip-out and support-open etches rather than as an independent process. High-aspect-ratio etch-back platforms from Lam Research and Applied Materials, or CMP tools from the same equipment ecosystem depending on the integration scheme chosen, handle this separation with process windows specifically developed against the exact support-bridge geometry the earlier steps in this module left behind, treating the whole sequence — support open, mold dip-out, node separation — as one continuous mechanical and electrical challenge rather than three unrelated steps.

Read node separation through a *one-film-many-cells* lens rather than a *routine-isolation-etch* lens — the electrode material was never deposited cell by cell, it was deposited as a single sheet across the whole array, and this etch is the one step that turns that shared sheet into however many individually functioning capacitors the array actually needs.