Mask Writing

Mask writing is the step where fractured design data actually becomes a physical pattern on the mask blank — a precision tool, either an electron beam or a laser, exposes resist on the blank shot by shot according to the fracture data, defining exactly where the pattern will later be etched into chrome. It's the one irreversible step in the entire mask-making flow: every upstream correction, every fracturing decision, converges here into a real, physical exposure that can't be undone once it's written.

Resist chemistry on the mask blank behaves the same conceptual way it does on a wafer, but the stakes are different — this single exposure defines the pattern every subsequent production wafer will inherit. Whichever writing technology exposes the resist, the chemically-amplified or conventional resist on the mask blank responds to that exposure exactly the way wafer resist responds to a scanner's exposure — latent image formation, post-exposure processing, development into the final pattern. The critical difference is scale of consequence: a wafer-level exposure error affects one wafer; a mask-writing error affects every single wafer that mask ever exposes for the rest of its working life.

Variable Shaped Beam e-beam writing remains the resolution benchmark against which every other mask-writing technology gets measured, which is exactly why it's still the dominant choice for leading-edge masks despite being purely serial. Two overlapping square apertures let the beam shape itself into a variable-sized rectangle for each individual shot, with shot sizes ranging from very fine, sub-micron dimensions up to several microns depending on the feature being written. Sub-nanometer placement accuracy, maintained through electrostatic beam correction and laser interferometry tracking the stage position continuously, is what lets VSB e-beam deliver the precision advanced mask patterns genuinely require — at the real cost of writing every single shot one after another, serially, which is exactly why advanced masks can take the better part of a full day to write.

Laser writing trades away VSB's fine resolution specifically to gain writing speed, which makes it the right tool for exactly the applications that don't need sub-30-nanometer mask precision in the first place. DUV laser exposure, sometimes combined with a programmable spatial-light-modulator mirror array for additional parallelism, can cover large mask areas considerably faster than a serial e-beam writer — but the fundamental resolution ceiling this technology hits, in the low hundreds of nanometers on the mask itself, simply isn't fine enough for the tightest features modern advanced-node OPC correction demands. This tradeoff is exactly why laser writing dominates mature-node masks, display photomasks, and MEMS applications while remaining essentially absent from leading-edge logic mask production.

Multi-beam writing's entire value proposition is refusing to accept the usual resolution-versus-throughput tradeoff at all — delivering e-beam-grade resolution while writing in a genuinely parallel, rather than serial, fashion. Hundreds of thousands of individually-controllable electron beamlets, each independently switched on or off, expose the mask simultaneously in a raster-scan pattern, achieving the same sub-10-nanometer resolution a single VSB beam delivers but completing the write in a fraction of the time a purely serial approach would require. This parallelism is also what makes multi-beam writing uniquely capable of handling curvilinear, non-Manhattan mask patterns (the kind inverse lithography optimization increasingly produces) without the shot-count explosion that would otherwise make writing such patterns with a serial e-beam writer impractically slow.

Choosing a mask-writing technology is ultimately a decision about what the specific mask actually needs, not a search for a single universally-best option, because resolution, throughput, and cost genuinely trade against each other differently across the three approaches. A leading-edge logic mask demanding the finest possible feature correction still often justifies VSB e-beam's serial write time, because resolution is non-negotiable at that level. A mature-node or display mask, where resolution requirements are comfortably loose, is better served by laser writing's dramatically faster throughput. And curvilinear EUV masks increasingly push toward multi-beam specifically because it's the only technology that avoids forcing a choice between the resolution those masks need and the throughput a production mask shop requires to stay economically viable.