multi-patterning
Light has a stubbornness built into physics: you cannot cleanly print lines finer than roughly the wavelength you print with allows. For years the industry pushed features far below that comfortable limit using 193nm light, and eventually the gap got so wide that a single exposure simply blurred adjacent lines into each other. Multi-patterning is the workaround. Instead of printing all the closely-spaced lines in one shot, you split them across two, three, or four separate exposures, each one printing a coarser, well-spaced subset that the optics can resolve. Picture painting a picket fence through a stencil whose slots are too fat to sit side by side: you paint every other picket, shift the stencil, and paint the ones in between. Two easy passes give you a fence twice as dense as either stencil could manage alone.
The catch is everything that hangs off the word 'separate.' Each exposure needs its own mask, and the self-aligned variants add rounds of deposition and etch to grow and trim sacrificial spacers that define where the real lines land. So one logical layer that EUV might print in a single step can balloon into two-, three-, or four-times the masks, plus the extra process steps between them. More masks mean more money, more time through the fab, and more chances for something to go wrong: every added step is another place for a particle, a misalignment, or an etch variation to creep in, and overlay error (how perfectly the second pattern lands on top of the first) becomes a yield-killer all its own.
This is precisely the limit EUV was built to relieve. Moving from 193nm light to EUV's 13.5nm wavelength shrinks the resolvable feature dramatically, so a layer that once demanded quadruple patterning can collapse back to a single EUV exposure. But EUV did not abolish multi-patterning, it just pushed it back: at the 5nm node and below, and on the very tightest layers even with EUV, fabs still reach for double patterning, and high-NA EUV will face the same tug-of-war on its first generation. Multi-patterning is the honest, expensive answer the industry kept giving while waiting for shorter-wavelength light to catch up to where the transistors had already gone.
Goal: print 4 dense lines (too tight for one 193nm shot)
Single EUV exposure Litho-etch double patterning (LELE)
-------------------- -----------------------------------
| | | | <- 1 mask mask A: | | | | (2 lines, spaced)
(13.5nm light mask B: | | (fill the gaps)
resolves it) result: | | | | (1 layer, 2 masks,
2 etches, overlay risk)One fine layer, two ways: EUV resolves the tight pitch in a single exposure; pre-EUV, you split it into two coarser exposures (each with its own mask and etch) that interleave to hit the same density.
It remains a live trade-off, not history: chipmakers constantly weigh the lower per-tool cost of multi-patterning with mature 193nm light against the fewer steps but pricier exposures of EUV, choosing layer by layer.