Pupil Fill Illumination Source Shape
# Pupil Fill and Illumination Source Shape
A lithography scanner's illumination doesn't have to fill its pupil uniformly, and for anything near the resolution limit, it shouldn't. Shaping the illumination into an off-axis pattern — a ring, a set of four poles, two poles — rather than a simple filled circle changes the angles at which light actually illuminates the mask, and that angular distribution directly controls how well the system resolves the specific pattern being printed. Pupil fill shaping is a lever process engineers pull before a single photon reaches the wafer, choosing illumination geometry to match the geometry of what's being patterned.
Off-axis illumination works because it controls which diffraction orders actually make it through the lens, and resolution depends on capturing enough of them. A mask pattern diffracts incoming light into a spread of orders, and the lens can only capture orders that fall within its numerical aperture; conventional, straight-on illumination centers the zero order in the lens's pupil and loses the higher orders carrying the pattern's fine detail first, which is exactly what limits resolution. Tilting the illumination off-axis shifts the whole diffraction pattern inside the pupil, and choosing that tilt angle correctly lets the lens capture the zero order and one first order simultaneously instead of two symmetric but weaker orders, improving both resolution and depth of focus for periodic features oriented to match the illumination's own geometry.
No single illumination shape is universally best, which is why source shape and mask layout get co-optimized together rather than chosen independently. Dipole illumination strongly favors lines running in one specific orientation, giving excellent resolution and depth of focus for that orientation while doing nothing useful, or actively hurting, lines running perpendicular to it — so a layout with both horizontal and vertical critical features often needs two separate exposures, each with dipole illumination rotated to match one orientation, rather than one exposure trying to serve both. Quadrupole illumination balances performance across both orientations in a single exposure at some cost relative to what dipole achieves for its one favored direction, and annular illumination serves dense, roughly isotropic patterns that don't have one dominant line direction to exploit. Source-mask optimization — co-designing the illumination pupil shape and the mask pattern together computationally, rather than picking a shape from a small fixed menu and hoping it works — has become a standard computational-lithography step for exactly this reason, run through the same ASML and Synopsys/Mentor-class software that also drives optical proximity correction.
Programmable illuminators, not fixed physical apertures, are what make this co-optimization practical at the pace advanced-node tape-outs demand. Early scanners used physical aperture plates to shape illumination, limiting source shape to whatever a small set of pre-machined apertures could produce and requiring a physical swap between layers needing different shapes. Modern scanners instead use programmable illumination systems — arrays of micro-mirrors or diffractive optical elements that can shape the pupil to essentially any desired distribution under software control — letting source-mask optimization software compute a custom, layer-specific illumination shape and have the scanner reproduce it without any physical hardware change between exposures.
Scanner dose uniformity (561) controls how much light reaches each point on the wafer. Pupil fill shaping controls something upstream and angular rather than magnitude-based — not how much light arrives, but from which directions — and getting that angular distribution right is what actually determines whether a pattern resolves at all before dose uniformity ever gets a chance to matter.