Abstract

As lithographic numerical apertures reach 0.5, we are halfway towards an ultimate limit of sorts. Unlike the once touted "1 μm barrier," the N.A. = 1 limit has physical meaning. Though still somewhat distant, its practical effect is to increase the cost of each incremental improvement in resolution. We will also face new image formation effects in the high-N.A. regime. Defocus tolerances will be tighter than the paraxial λ/(N.A.)² scaling. Polarization in lithographic printers will cause image asymmetries. Rising N.A. costs stimulate interest in strategic manipulation of image formation within constrained N.A. Most proposals involve removal of light from the center of the pupil, producing an angular spectrum more concentrated in oblique k_z, components. This improves focal tolerances by reducing relative dephasing during z propagation. Edge slopes are also sharpened, at the cost of reduced image fidelity due to accentuated bipolar amplitude lobes. "Soft image manipulation techniques" attempt to hold these lobes below the resist threshold. "Hard techniques" attempt instead to precisely match positive and negative lobes, if pattern topology permits, thereby exploiting the 2× higher bandwidth of the intensity image over the amplitude image. The phase tilt from oblique illumination can also provide approximate positive/negative phase coloring, if additional negative phase is added to balance positive and negative lobes. Segmentation of the pattern in separate exposures can provide needed flexibility. All such techniques possess numerous practical difficulties. Moreover, there is a limit to the image sharpness that can be produced from fixed N.A. There is no full substitute for continued improvements in lens resolution, despite the increasing cost.

© 1992 Optical Society of America