Mandrel Etch

The mandrel etch, also called core etch, cuts the sacrificial core lines that self-aligned double and quadruple patterning build the rest of a DRAM array's dense pitch around — the mandrel itself never survives to the finished chip, but its exact width and spacing is what determines the width and spacing of every real feature that gets defined relative to it later. DRAM cell pitch has shrunk well past what a single lithography exposure can resolve on its own, so instead of trying to print the final fine pitch directly, the process prints a coarser, lithography-friendly mandrel pattern first and then uses it purely as a scaffold.

The trick that makes this whole scheme work is that the mandrel contributes nothing to the final pattern except its edges. Once the spacer material is deposited conformally around every mandrel line and etched back to leave only vertical sidewall spacers standing, the mandrel's job is finished — the spacers that remain, set by deposition thickness rather than by lithography resolution, are what actually define the fine final pitch, and a repeated spacer-on-spacer version of the same trick (self-aligned quadruple patterning, or SAQP) pushes the achievable pitch even finer for a given starting lithography resolution.

Because the mandrel will be deliberately removed later, this etch gets judged almost entirely on sidewall profile rather than on the mandrel's own survival. A mandrel with a sloped, rounded, or rough sidewall transfers that same imperfection directly into the spacer deposited around it, and that spacer is the structure that actually survives into the finished chip — a mandrel etch with excellent critical-dimension control but a poor sidewall angle can still ruin the pattern it was supposed to enable, which is why this etch is tuned as carefully for profile straightness as for linewidth itself.

Core removal, the mandrel-pull step that follows, has to clear the mandrel material completely without disturbing the spacers it leaves behind, and that selectivity requirement shapes what mandrel material gets chosen in the first place. Amorphous carbon and polysilicon mandrels are both common choices specifically because each has an etch or ash chemistry that removes it cleanly while leaving an oxide or nitride spacer essentially untouched; DRAM makers tune this mandrel-etch-and-pull pair as a matched set on the same high-resolution plasma etch platforms — Lam Research and Tokyo Electron among the primary suppliers — used throughout the rest of the cell module's patterning sequence.

Read the mandrel etch through a *scaffold-not-structure* lens rather than a *just-another-line-etch* lens — getting the mandrel's own dimensions exactly right matters far less than getting its sidewall shape right, because everything the chip actually keeps is defined by the spacer wrapped around a line that, by design, won't be there by the time the wafer is finished.