Seung-yoon Lee

Properties of EUVL masks as a function of capping layer and absorber stack structures

We have fabricated extreme ultraviolet lithography (EUVL) blank masks consisting of a TaN absorber, Ru capping layer, and Mo/Si multilayers using ion-beam sputter deposition and investigated their dependence on capping layer and absorber stack structure. At EUV wavelengths, the reflectivities of the multilayers, including their dependency on the thickness of the capping and absorber layers, are in good agreement with simulation results obtained using Maxwell equations and the refractive indexes of each layer. Ru, one of the most promising capping materials on Mo/Si multilayers due to its resistance to oxidation and selectivity to etching, also shows better EUV reflectivity than Si as a capping layer if we choose a thickness that produces a constructive interference. To meet the reflectivity requirements (⩽ 0.5 %) in the SEMI EUVL mask standard specifications, a TaN absorber at least 70 nm thick should be applied. However, aerial image results simulated by using EM-Suite show that 40 nm is sufficient for the TaN absorber to display the maximum image contrast. In addition, horizontal-vertical (HV) biasing effects due to mask shadowing become negligible if the TaN is reduced to about 40 nm. As a result, we suggest using a thin TaN absorber 40 nm thick since it is able to minimize mask shadowing effects without a loss of image contrast.

Investigation on accuracy of process overlay measurement

The shrinkage of design rule necessitated corresponding tighter overlay control. However, in advanced applications, the extension of current technology may not be able to meet the control requirement, consequently, additional breakthroughs are required. In this study, we investigated methods to enhance the overlay control, approaches by extraction of real overlay error out of overlay measurement. So far, only the destructive inspections like vertical SEM have enabled us to measure real misalignment. But, a concept of non-destructive method is proposed in this paper, extracting vertical information from the results of multiple measurements with various measurement conditions, keys or recipes. With this proposed method, the measurement accuracy can be improved and we can enable a new knob for overlay control.

Smart overlay metrology pairing adaptive deep learning with the physics based models used by a lithographic apparatus

All wafers moving through a microchip nanofabrication process pass through a lithographic apparatus for most, if not all, layers. With a lithographic apparatus providing a massive amount of data per wafer, this paper will outline how physicsbased models can be used to refine UVLS (ultraviolet level sensor) metrology into four unique inputs for use in a deep learning network. Due to the multi-dimensional cross correlation of our deep learning network, we then show that training to a sparse overlay layout with dense inputs results in a hyper dense overly signature. On a testing dataset blind to the training we show that the accuracy of the predictive computational overlay metrology can capture R2 up to 0.81 of the signature in overlay Y. As a real-world application, we outline how our predictive computational overlay metrology can then be used to designate which wafer combinations, coming from the TWINSCAN system, should have overlay measured with a YieldStar system for possible use with APC (advanced process control).

The effect of individually-induced processes on image-based overlay and diffraction-based overlay

In this paper, set of wafers with separated processes was prepared and overlay measurement result was compared in two methods; IBO and DBO. Based on the experimental result, theoretical approach of relationship between overlay mark deformation and overlay variation is presented. Moreover, overlay reading simulation was used in verification and prediction of overlay variation due to deformation of overlay mark caused by induced processes. Through this study, understanding of individual process effects on overlay measurement error is given. Additionally, guideline of selecting proper overlay measurement scheme for specific layer is presented.

Feasibility Study of Matched Machine Overlay Enhancement toward Next Generation Device Development

In this study, we proposed the concept of high order field-by-field correction for Matched Machine Overlay (MMO) error minimization and we have validated it through experiments. Because scanners have unique grid fingerprint, MMO value between machines is higher than the one of Single Machine Overlay (SMO). In some cases, the localized grid distortion mainly contributes to the MMO value. However, this localized grid distortion cannot be flatten by a normal correction method such as 10-parameter correction. Until now, in order to flat the localized grid distortion, ultimate correction capability can be realized by combining 6-parameter field-by-field correction and intra-field high order correction methods. However 6-parameter could be not enough to follow the diversity of local distortion. In this study, for further improvement of MMO, high order field-by-field correction capability was investigated and the results were compared. Base on simulation, we found that the field-by-field correction was a successful way to lower the MMO value of EUV vs. ArF immersion scanners. By experimental demonstration, it showed that field-by-field correction was more effective to correct localized grid distortion and the gain via high order model was about 0.5 nm. These results will be helpful to achieve the MMO specification for the next generation device.