Microfabrication has developed rapidly in the last few decades. As in all areas of engineering, simulation of physical processes is becoming more important. Process optimization through optical lithography simulation, a well-established tool, reduces production costs and time-to-market. To simulate light propagation and photoresist behavior, it is necessary to characterize the resist as it is developed. To obtain this information, the thickness of transparent and semi-transparent films has to be measured rapidly and non-destructively.
Much information can be obtained by measuring the thickness reduction of the photoresist over time as it is developed. This information can be calculated from measurements of reflectivity versus wavelength. When measured for several different exposure intensities, the results can also be used to calibrate models of photoresists in software simulations.
Groups at different institutions have designed a number of different laboratory setups to analyze dissolution rates of photoresists. Thickness information can be calculated either by measuring reflectance of a single wavelength or a spectrum of wavelengths. Most multi-wavelength approaches are based on least squares fitting of theoretical to measured reflectance, using assumed values for the refractive indices.1–3 A simpler method, based on determining the location of the peaks within the spectrum,4 is advantageous because it would require less sophisticated hardware. Both single and multiwavelength approaches can only provide relative thickness information, and are based on the positions of peaks in the spectrum. For thin resist layers, no peaks are visible, so the resist has to be developed completely before absolute thickness can be calculated. This makes it difficult to characterize photoresist dissolution behavior.
Our approach combines least squares fitting and the simpler optical inspection method. The mathematical analysis we use provides good accuracy as long as some restrictions concerning the optical constants of the components used (such as substrate, photoresist and developer) are fulfilled. Thus, characterization of the development process of a large variety of photoresists, including diazonaphthoquinone-Novolac-based variants as well as of chemically-amplified resists, is possible with our approach.
Figure 1. Setup of the dissolution rate monitor which shows developer feed, developer cell, light source, spectrometer, and software for data acquisition and data analysis.
There are several different development methods. The most commonly-used version is a single-puddle development, which we also chose for our dissolution rate monitor (DRM) development test-bench (see Figure 1). The data record starts when the developer arrives between the probe (optical fiber) and photoresist, which results in a characteristic peak in the reflectivity versus time diagram. A schematic view of our setup is shown in Figure 2. Depending on the exposure dose, the dissolution rate of the resist changes. At low doses, the development process does not completely remove the resist from the substrate surface. In this case, the absolute thickness values are necessary to calculate the complete thickness values over time. For thin resist layers (<500nm), it is difficult to measure the absolute thickness value, so we developed an advanced normalization method to improve the accuracy of the results, which leads to accuracy down to ∼100nm.
Figure 2. Schematic view of the dissolution rate measurement during the development step. (a) Fiber-optic probe, exposed resist, and substrate when the developer covers the resist. (b) Resist erosion. (c) The point when no exposed resist remains.
Figure 3. (a) Repeatability test for a diazonaphthoquinone-based AZ 1518 resist. (b) Final dissolution rate monitor for AZ 6624 at different exposure intensities.
Our data acquisition provides reflectivity information at time intervals of 10ms and longer, resulting in thickness versus time information with a corresponding resolution. Typical dissolution rates of photoresists range between 100–400nm per second. Thus, phenomena such as standing waves, and the nonlinear progression of dissolution, can be measured accurately. During the verification of the DRM, we found that a resolution of 1nm is sufficient. Figures 3a and b show a repeatability curve and a final DRM for AZ6624 at different exposure doses, respectively. The accuracy of the absolute thickness versus time information provided by the DRM depends on the accuracies of the refractive index assumed for the resist and of the optical model. We developed mathematical models to estimate contributing factors to accuracy and to estimate error propagation. Thus, the maximum possible absolute error can be calculated each time a spectrum is collected, which defines a confidence interval with regard to the thickness versus time information obtained. The availability of such a small confidence interval is unique to DRM.
Figure 4. (a) Micrograph of a 500nm-wide resist structure of SPR955-CM. (b) Simulation of the resist structure using data from the dissolution rate monitor.
Once thickness values were established, we used the DRM data as an input parameter for resist calibration in simulation software, such as LayoutLAB (GenISys) or Dr. LiTHO (Fraunhofer IISB). Figure 4 shows a comparison of a real 500nm structure and a simulated one. The standing waves, which are characteristic for the resist when using no bottom anti-reflective coating layer, are exactly reproduced.5
The simplicity of our approach affords both increased flexibility and expands on the photoresist characteristics that can be determined, as well as providing accurate absolute thickness information. In addition, our analytical approach yields a confidence interval, giving the user an idea about the resulting accuracy. Our data analysis technique provides a robust in-situ method for inspecting the development of photoresists for a wide range of thicknesses, yielding a significant improvement over previous approaches. Our next steps are to adapt the development cell to different development methods, including multi-puddle and immersion.
We would like to thank the Bayerische Forschungsstiftung for partially funding this work in the framework of the Mask-Aligner Lithography Simulation project.
Microtechnology Research Center
University of Applied Sciences
Department of Microsystems Engineering (IMTEK)
University of Freiburg, Germany
Microtechnology Research Center
University of Applied Sciences
Integrated Systems and Device Technology
1. T. Benson, L. Kamlet, P. Klimecky, F. Terry, In-situ spectroscopic reflectometry for polycrystalline silicon thin film etch rate determination during reactive ion etching, J. Electron. Mater. 25, p. 955-964, 1996.
2. W. K. Ho, X. Q. Chen, A. Tay, X. D. Wu, In-situ monitoring of resist thickness contour in microlithography (Volume 7 No. 4), Tech. Rep. 7(4), 2006.
3. Y. K. Sin, K. Ibrahim, Y. O. Kok, A. Aziz Abdul, K. P. Lim, Film thickness determination by MATLAB assisted ellipsometry, Solid State Sci. Technol. 13, p. 143-150, 2005.
4. C. L. Henderson, Advances in Photoresist Characterization and Lithography Simulation, 1998.
5. S. Partel, Fabrication process development for a high sensitive electrochemical IDA sensor, Presented at MNE 2011 Berlin, 2011.