Calibration of physical resist models for simulation of extreme ultraviolet lithography

摘要

We have calibrated a physical resist model for extreme ultra-nviolet (EUV) lithography, and discuss model calibration and validation overna larger set of structures. The study is conducted on an extensive datanset, collected at imec, for ShinEtsu resist SEVR-59 exposed on the ASMLnEUV alpha demo tool (ADT). The data set included more than a thousandnmeasured feature widths (critical dimensions or CD) on wafer and mask.nWe address practical aspects of the calibration, such as the speed ofncalibration and selection of calibration input. The model is calibrated bynsimultaneously fitting 12 process windows of features with different masknCD (32, 36, 40 nm), orientation (horizontal, vertical), and pitch (dense,nisolated). The smallest feature size at nominal process condition is an32 nm CD at a dense pitch of 64 nm. Mask CD metrology was used tonfit the model versus actually measured mask CD’s. Cross-sectional scan-nning electron microscopy information was included in the calibration, tontune the simulated resist loss and sidewall angle. The achieved calibrationnroot-mean-squared (RMS) error is ∼1.0 nm.We discuss the elements thatnwere essential to obtain a well calibrated model.We discuss the impact ofn3-Dmask effects on the Bossung tilt.We demonstrate that a correct repre-nsentation of the flare level during the calibration is key in order to achievena high CD predictability at various flare levels. Although the model calibra-ntion is performed on a limited subset of the measurement data collectednon 12 different patterns (one dimensional structure process windows),nits accuracy is validated on a large number of patterns used to calibratenmodels for optical proximity correction—several hundred different featurentypes, at nominal dose and focus conditions. These were not included innthe calibration; validation RMS results as small as 1 nm can be reached.nFurthermore, we study the model’s extendibility to two-dimensional end ofnline structures. Finally, we show that we can correlate the experimentallynobserved fingerprint of the EUV ADT CD uniformity, to a CD fingerprintncalculated using this resist model, where EUV tool and mask specificnsignatures are taken into account