Spin-on Carbon Process

The spin-on carbon process applies SOC the same way the spin-on glass process applies SOG — as a liquid, centrifugally spread and baked — but SOC's modern job is different from either SOG's dielectric gap-fill role or the plasma-deposited amorphous carbon hard mask covered in the carbon and APF processes. SOC's primary use today is as the thick bottom layer of a trilayer lithography resist stack, a scheme that exists because ordinary single-layer photoresist can't simultaneously resolve the finest features a scanner can print and survive the deep, aggressive etch that pattern then has to drive into whatever's underneath.

Spin-coat SOC
Thick planarizing underlayer, liquid applied
Bake / cross-link
Thermally cures, won't re-dissolve in later coats
Trilayer resist stackeach layer does one job the others can't
Si-ARC middle layer
Optical ARC + etch-transfer mask
Thin photoresist on top
Only has to resolve the pattern, not survive the etch
Pattern transfers down through Si-ARC
Into the thick SOC layer
SOC survives the deep final etch
Protects what's underneath, planarized from the start

A single resist layer faces two contradictory requirements as features shrink and topography gets rougher, and a trilayer stack exists specifically to split those requirements across three separate films instead of asking one film to satisfy both. Resolving the finest features a lithography tool can print wants a thin resist film, because thick resist diffracts and defocuses the exposure light enough to blur fine patterns — the same optical constraint the APF process covers for a different hard-mask context. But surviving a deep anisotropic etch down into whatever layer sits far below the resist wants a thick, etch-resistant film, the opposite requirement entirely. SOC resolves that contradiction the same way amorphous carbon resolves it for a single-layer hard mask: let the thin top layer handle resolution only, and let a much thicker underlying film absorb the etch.

The Trilayer Resist Stack each layer does exactly one job — resolution, optical control, or etch survival — never more than one Cross-section, top to bottom Photoresist — thin, resolves the pattern only Si-ARC — optical control + etch mask SOC — thick, planarizing absorbs the deep final etch fills severe topography variation Target layer / rough topography resist only has to survive the exposure and a shallow etch SOC absorbs the real etch — resist is already gone by then this is the same thin-resist/thick-sacrificial-layer logic the APF process uses — split across three layers instead of two

SOC's liquid-spin application is what lets it planarize severe topography variation in a way a conformal plasma-deposited film can't match, which is exactly why it's chosen over a plasma amorphous carbon film for this specific underlayer role. A chip's surface rarely presents uniform topography — dense memory arrays sit next to open logic areas, deep trenches sit next to flat field regions — and a conformal deposition follows all of that variation rather than flattening it. SOC, applied as a liquid the same way spin-on glass is, self-levels across that variation during the spin-coat step itself, giving the trilayer stack a genuinely flat top surface for the Si-ARC and photoresist layers to sit on regardless of how rough the real topography underneath happens to be — planarization the whole stack depends on, since lithography exposure quality degrades badly over any significant surface unevenness.

The silicon-containing middle layer earns its place by doing two jobs that would otherwise need two separate films: optical anti-reflective control for the exposure, and a hard mask for transferring the pattern down into the SOC. Silicon content in that middle layer gives it strong etch selectivity against the carbon-based SOC below it — the same selectivity principle the APF process uses between amorphous carbon and its cap layer — so a fluorine-based etch can pattern the Si-ARC precisely without touching the SOC, and then a completely different, oxygen-based etch chemistry (which attacks carbon readily but barely touches silicon-containing films) transfers that same pattern down through the thick SOC layer underneath.

Once pattern transfer reaches the SOC, the photoresist and Si-ARC layers have both already done their jobs and are typically consumed or irrelevant — the SOC alone carries the pattern through the deep final etch into the real target layer. That division of labor is the entire payoff of the trilayer approach: no single film in the stack ever has to be both thin enough to resolve the finest pitch a scanner can print and thick enough to survive the deepest etch a target layer demands, because those two requirements were never asked of the same material in the first place.

Read the spin-on carbon process as the point where two separate ideas covered elsewhere in this glossary converge into one modern lithography scheme: SOG's liquid-spin planarization mechanism, applied to carbon instead of glass, gives a trilayer resist stack the flat starting surface it needs across uneven topography, while the thin-resist/thick-sacrificial-underlayer logic from the APF process gets split a second time, across a dedicated optical/etch-mask middle layer, so that resolving a pattern and surviving the etch that pattern eventually drives are never demands placed on the same film at once.

Take spin-on carbon process further

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