Simulation and evolutionary optimization of electron-beam lithography with genetic and simplex-downhill algorithms

Abstract

Genetic and simplex-downhill (SD) algorithms were used for the optimization of the electron-beam lithography (EBL) step in the fabrication of microwave electronic circuits. The definition of submicrometer structures involves complex exposure patterns that are cumbersome to determine experimentally and very difficult to optimize with linear search algorithms due to the high dimensionality of the search space. A SD algorithm was first used to solve the optimization problem. The large number of parameters and the complex topology of the search space proved too difficult for this algorithm, which could not yield satisfactory patterns. A hybrid approach using genetic algorithms (GAs) for global search, and a SD algorithm for further local optimization, was unable to drastically improve the structures optimized with GAs alone. A carefully studied fitness function was used. It contains mechanisms for reduced dependence on process tolerances. Several methods were studied for the selection, crossover, mutation, and reinsertion operators. The GA was used to predict scanning patterns for 100-nm T-gates and gate profiles with asymmetric recess and the structures were fabricated successfully. The simulation and optimization tool can help shorten response times to alterations of the EBL process by suppressing time-consuming experimental trial-and-error steps.

High resolution electron beam lithography of PMGI using solvent developers

“…While it is generally believed that PMGI will not be dissolved by solvent developers [3,5], in this paper we will show that, at higher dose (>500 lC/cm 2 for 30 kV), the common solvent developers for PMMA can actually be used to develop PMGI with a performance similar to that of PMMA. PMGI displays several advantages over PMMA: resistant to general solvents, less susceptible to intermix with other polymers or resists when stacked [2], slightly more resistant to acids, slower dry etch rate (by 30% in CHF 3 plasma), and higher thermal stability with a glass transition temperature of 189°C (vs. 105°C for PMMA).…”

Analysis of Fitness Functions for Electron-Beam Lithography Simulation and Evolutionary Optimization

“…In this letter, we extend the work presented in [4], and study the limitations of the fitness function F eucl: dist: based on E 2 D ( ), which gives rise to some small wiggles and dips in the optimized profile, as is shown in Fig. 1.…”

“…In a previous paper [4], we reported on the development of an EBL simulation and optimization tool and its application to the manufacturing of T-shaped gates of high-electron mobility transistors. Given an ideal resist profile to be fabricated with EBL, this tool finds the spatial distribution of exposure times (i.e., the scanned pattern) that generates this profile.…”

“…This computed profile is calculated as a function of the electron dose, the proximity parameters, and the local resist solubility. In [4], the fitness function F eucl: dist: was defined as the inverse value of the square of the Euclidian distance E 2 D ( ) between the ideal profile and the computed profile where (x t ; z t ) are the coordinates of the nodes along the chain of the target profile, (xcomp; zcomp) are the coordinates of the nodes along the chain of the computed profile for the scanned pattern , w(k) is the weighting factor of node k, and N is the number of nodes along both chains.…”

Fundamentos, Potencialidades e Aplicações de Algoritmos Evolutivos

“…Esta concepção de método simplex pode ser utilizada em diversas variantes possíveis de AEs híbridos [139], [175], [133].…”