Mr. Abdullah Karatas Profile

Abdullah Karatas

at Rheinland-Pfälzische Technische Univ. Kaiserslautern-Landau

SPIE Involvement:

Author

Publications (2)

Proceedings Article | 21 June 2019 Presentation + Paper

Development of a convolutional autoencoder using deep neuronal networks for defect detection and generating ideal references for cutting edges

Abdullah Karatas, Dorothea Kölsch, Samuel Schmidt, Matthias Eifler, Jörg Seewig

Proceedings Volume 11056, 1105623 (2019) https://doi.org/10.1117/12.2525882

KEYWORDS: Neural networks, Convolution, Defect detection, Machine learning, Data modeling, Gaussian filters

Read Abstract +

Proceedings Article | 14 July 2017 Paper

Application of GPU-based, highly parallelized algorithms for the estimation of electromagnetic multi-layer interactions

François Torner, Abdullah Karatas, Matthias Eifler, Indek Raid, Jörg Seewig

Proceedings Volume 10329, 1032922 (2017) https://doi.org/10.1117/12.2270257

KEYWORDS: Electromagnetism, Oxidation, Optical coatings, Statistical modeling, Ellipsometry, Graphics processing units, Optical properties, Aluminum, Oxides, Multilayers, Refractive index, Polarization, Phase shift keying, Silicon

Read Abstract +

In nowadays industry, complex surfaces with material contrasts or surface coatings are present and represent a challenge for optical topography measuring instruments. The reason is that varying optical properties lead to phase jumps and to topography deviations when the surface height is evaluated. Thus, Ellipso-Height-Topometry as a measurement technique which can measure both topography and material properties of technical surfaces was proposed in order to achieve a correction of the occurring topographic artefacts. The height correction value can be obtained for the compensation of material-induced height deviations and the thickness of surface layers can be evaluated. Currently, it is possible to calculate the surface characteristics from ellipsometric parameters for at most two layers. However, the described height corrections are only possible when well-defined and realistic models of surface layers can be utilized, e.g. a given set of homogeneous oxide layers. Oxidation effects however describe statistical processes which can be predicted with underlying material distribution models. This leads to an uncertainty in ellipsometry, which is considered with a new approach that will be discussed in this publication. Therefore, an extended multi-layer approach which is capable of handling additional layers based on a parallelized algorithm using graphic processing units and the commonly known CUDA technology is proposed. This algorithm can also be used to consider material proportions which result from oxidation effects in z direction. The new approach for the Ellipso-Height-topometry measurement technique is compared with the current procedures which often neglect the existence of an oxide layer for the basic material. To experimentally verify the approach and according algorithm, it is applied for the evaluation of actual surfaces with multiple plane layers and different materials. Test samples with different materials are used in order to evaluate the complex refractive index, the distribution of identified materials and the layer thicknesses with actual Ellipso-Height- Topometry measurements. The results of the measurements are compared to the predicted theoretical results.

View contact details

UPDATE YOUR PROFILE

Is this your profile? Update it now.

Sign into your SPIE.org account

Don’t have a profile and want one?

Create an account on SPIE.org

Keywords/Phrases

Search In:

Publication Years