Effects of mathematical models and algorithms on quantitative characterization of areal step height with optical and stylus profilometers

Step height characterization is essential for the quality control of various functional components, such as graphene and the step features of semiconductor devices. Two methods are proposed to characterize the areal step heights. The first method extends the two-dimensional characterization in the I...

Full description

Saved in:
Bibliographic Details
Published inPrecision engineering Vol. 72; pp. 777 - 788
Main Authors Wang, Chen, Yu, Yingjie, Zhang, Xiangchao, D'Amato, Roberto, Gomez, Emilio
Format Journal Article
LanguageEnglish
Published Elsevier Inc 01.11.2021
Subjects
Online AccessGet full text
ISSN0141-6359
1873-2372
DOI10.1016/j.precisioneng.2021.08.007

Cover

More Information
Summary:Step height characterization is essential for the quality control of various functional components, such as graphene and the step features of semiconductor devices. Two methods are proposed to characterize the areal step heights. The first method extends the two-dimensional characterization in the ISO specification into a three-dimensional one by extracting multiple parallel profiles. The second method calculates the step heights by projecting from the measurement points to the normal vector at the surface centroid. Mathematical models and algorithms of the two methods are introduced and validated by synthetic data. Experiments are conducted by comparing the assessment results of the two methods and of a method proposed in a previous research. The calibrated values of the standards are utilized for validation. The characterization results may differ notably or slightly, depending on the properties of the data and the algorithm. •An ISO based method for characterizing 3D step heights is proposed.•A Centroid-Normal method for characterizing 3D step heights is proposed.•The same evaluation data interpret different parameterization with the two methods.•Discrepancies can be notable or slight depend on the data properties.•Discrepancies of the results come from residues of Least Square fitting.
ISSN:0141-6359
1873-2372
DOI:10.1016/j.precisioneng.2021.08.007