<title>Photoresist metrology based on light scattering</title>

Abstract

Angle resolved light scatteromeiry along with advanced data analysis is a promising new metrology technique to meet the challenges oftoday's and tomorrow's submicron technology. The measurement accuracy strongly depends on the performance capabilities of the algorithms utilized for data exploration and analysis. Presently, multivariate regression methods such as inverse least squares and principal component approaches are prefened. Substantial accuracy gains may be achieved by applying quasi-nonlinear methods, i.e. nonlinear data pretreatment followed by the usual linear regression. In this way, not only were the linewidth prediction errors in measuring developed resist lines pushed to below 20 nm, but likewise more complicated tasks as silylation profile evaluation and latent image measurement could be addressed satisfactorily. . INTRODUCTIONRecently, a new method was introduced in micrometrology, which consists in a combination of angle resolved scattered light measurement (ARS) and sophisticated data analysis '. A high measurement throughput, sufficient accuracy in the submicron range, nondesiructive probing and on-line capability are the crucial benefits ofthis metrology. The basic principle ofthe light measurement is straightforward: A focused laser beam is impinging on the surface under investigation. Caused by the diffraction and the microroughness ofthe sample, the light is scattered into the full hemisphere above the specimen. Considering only onedimensional surfaces (e.g. line space patterns) and adjusting the incident beam perpendicular to the lines, the light distribution will essentially be confmed to the plane of incidence. The resulting light distribution strongly depends on the topographic and optical properties ofthe sample. Thus, it may be regarded as a kind of "optical fingerprint". Particularly, ifa periodic line/space pattern is illuminated with a laser spot covering more than a few pitches, sharp diffraction orders stand out from the noisy background. Now, the difficulty remains of identifying the profile ofthe scatterer from this "fingerprint". Due to the ambiguity between scattering surface and measured light intensity distribution, an analytical solution ofthe problem may not be gained in general. However, if certain a-priori information about the surface profile and the refraction indices is given, a quantitative assignment between light intensities and profile parameters such as linewidth, height, slope angle and others may be obtained by establishing multivariate regression models. Until now, several applications such as DRAM trench depth characterization 2, latent image and post exposure bake monitoring , PSM meirology and submicron linewidth resist metrology based on scatterometry were published. Predominantly, linear regression models including the inverse least squares method (ILS), the principal component regression (PCR) and the partial least squares method (PLS) known from chemometrics 6 are applied. Owing to the linearization inherent in these models, the resulting accuracy is ...

Scatterometry—fast and robust measurements of nano-textured surfaces

“…For gratings with a pitch comparable to the wavelength of the incoming light, it can be observed from equation (5) that at most a few diffraction orders exist. Instead, by only studying the specular reflection as a function of the angle of incidence, it is possible to measure sub-wavelength gratings [54][55][56][57][58]. In the measurements of the specular reflection, the detector is always positioned symmetrically at the normal of the substrate with respect to the incoming light, see figure 4(B), and hence the name 2-theta (2θ).…”

“…In the measurements of the specular reflection, the detector is always positioned symmetrically at the normal of the substrate with respect to the incoming light, see figure 4(B), and hence the name 2-theta (2θ). Originally, the 2-theta scatterometer was demonstrated for overlay accuracy measurement [55], critical dimensions and grating profile [54,57], gratings on multilayers [56] and combined CD and overlay measurements [58]. Later the technique was used to characterize the profile parameters for 2D patterns with holes and dots in resist layers [59] and the linewidth roughness of a grating [60].…”

Characterization of optical diffraction gratings by use of a neural method

“…Another method consists of retrieving the profile parameters by data analysis based on different statistical models, for instance, classical linear regression models, such as the principal components analysis (PCA) method, 2 the discriminant analysis method, 3 and the partial-least-square method. [4][5][6] In the past few years another statistical tool has received increased attention: the neural network. A neural network analysis can be regarded as a regression method based on a non-linear model.…”