Source optimization method by adopting self-adaptive compressed sensing technology

Abstract

The invention discloses a source optimization (SO for short) method by adopting a self-adaptive compressive sensing (CS for short) technology. Compared with an existing CS-SO algorithm, observation point distribution obtained by the SO method in the invention can better represent topological structure characteristics of a circuit layout; on the condition of a same number of observation points, computation efficiency is higher; on the condition of similar computation time of the algorithm, imaging performance of a laser lithographic system can be further improved. According to the source optimization (SO for short) method by adopting the self-adaptive compressed sensing (CS for short) technology, a blue noise sampling method is adopted to select the observation points on the circuit layout, and with a constraint condition that an imaging value at each observation point is equal to a target imaging value, a constraint condition system of linear equations is constructed. Afterwards, a self-adaptive projection matrix is constructed by using information of the circuit layout. According to the CS theory, the self-adaptive projection matrix is adopted to compress the dimensionality of the constraint condition system of linear equations, and an SO optimization question is transformed to an image restoration question of solving L-p (0<=p<=1) norm, and a CS signal is adopted to reconstruct an algorithm to optimize a source image.

Description

technical field [0001] The present invention provides a source optimization (SO) method using adaptive compressed sensing (compressive sensing, CS) technology, which belongs to the technical field of lithographic resolution enhancement. Background technique [0002] The lithography system is the core equipment used to manufacture VLSI with micron-scale and nanoscale line width. The lithography system is mainly composed of four parts: an illumination system, a mask, a projection objective lens, and a silicon wafer coated with photoresist on the upper surface. The light emitted by the light source is irradiated and passes through the mask to form a diffraction field. An image of the mask pattern is formed on the silicon wafer. [0003] For the currently commonly used 193nm ArF deep ultraviolet lithography system, as the lithography technology node enters 45-14nm, the critical dimensions of integrated circuits have entered the deep sub-wavelength range. At this time, the resol...

AI Technical Summary

Problems solved by technology

Therefore, when there are many observation points, the computational efficiency of the algorithm is significantly reduced; and when there are few observation points, the light source optimization result obtained by the algorithm will deviate from the optimal solution of the SO problem, affecting the imaging performance of the optimized lithography system

[0007] Second, the above method adopts a random sampling method when selecting observation points, so it is likely that the distribution of observation points is not uniform (that is, some areas of the circuit layout concentrate a large number of observation points, while other areas contain only a small number of observation points). Observation points, or even no observation points), at this time, some topological feature structures on the circuit layout cannot be represented by the distribution of these observation points, so it is easy to lose key information in the circuit layout structure, resulting in the obtained light source optimization results Deviation from optimal solution for SO problem

[0008] In summary, the existing SO methods need to be further improved in terms of approaching the optimal solution of the SO problem, the selection of observation points, and the efficiency of algorithm operations.

Images