Model-based SEM for dimensional metrology tasks in semiconductor and mask industry

Abstract

The requirements on the use of scanning electron microscopy (SEM) as a measurement technique for the process control of dimensional parameters are most challenging in the production of masks and wafers in the semiconductor industry. Subtle details of SEM signal variations have to be thoroughly understood to be able to monitor and trace small process variations to changes in dimensional parameters of the features to be controlled. This paper reviews the fundamentals, special performance features and applications of existing SEM image contrast simulation packages based on Monte Carlo methods.

Challenges in nanometrology: high precision measurement of position and size

“…In order to take into account the stochastic nature of the different elastic and inelastic scattering events of the primary electrons as well as the resulting electrons (backscattered, secondary or transmitted electrons) in the solid state the majority of them uses Monte Carlo simulations. An overview of the applied models is provided in [26], and a detailed description of MCSEM, the modular Monte Carlo software package for simulation of electron signal contrast developed at the PTB, has been published in [27]. Uncertainties of CD-SEM measurements at PTB for the width of high quality lithographic features of sizes down to 100 nm and below on photomasks and wafers as low as U 95% = 10 nm have been reached.…”

Scatterometry reference standards to improve tool matching and traceability in lithographical nanomanufacturing

“…The top down SEM images shows rather smooth line edges. Applying PTB's edge detection algorithms [20], these top down images are analyzed to derive the local line edge roughness and to some extend also local line edge angles.…”

Monte Carlo Simulation of SEM and SAM Images

“…Frase et al have recently reviewed the fundamentals, special performance features and applications of existing SEM image contrast simulation packages based on MC methods (Frase et al, 2009). It is considered that the MBL method should be the most accurate algorithm for its sound physical foundation and one-to-one relationship between the image intensity profile and the geometry models (Villarrubia et al, 2005a).…”

“…In Ref. (Frase et al, 2009), Frase et al also reviewed the physics of probe-sample interaction and modeling, and existing MC simulation programs. Both of them found that, the MC model based on the Mott's cross-section for elastic scattering and a dielectric function approach for inelastic scattering would achieve the most agreeable result comparing with experiment.…”