Lithography Simulations for an Ultratech Stepper

Summary and conclusion: The manufacturing of structure in the μm regime with lithography tools can be realized very fast with different techniques in these days, but the analysis of the structures (CD-SEM) is very time consuming. Lithographic modeling is used to simulate the lithographic process and to calibrate different models (like photoresist models) what saves much time and costs.

So this work was concerned with assembling an easy to use calibration progrm for DNQ type photoresits calibration using exposures by a stepper. This was achieved with the use of Python which is utilized as scripting language for the Dr.Litho lithographic modeling tool from Fraunhofer IISB.

The exposure experiments were done with the Ultratech Sapphire 100E stepper in the clean room of the Fraunhofer IISB as well as the CD measurements with a SEM. The AZ5214E photoresist was used for coating the wafers. This so called image reversal resist can be used for both, application of a positive tone and negative tone photoresist due to the processing of it. So first it was applied as a positive photoresist and exposure was made with a dose variation in range of 130mJ/cm2 - 200mJ/cm2 for mask linewidth feature widths of 1ηm and 2 ηm. The beavior of the measured CDs was not distinct enough to apply a calibration so the calibration procedure explained in chapter 6.2.1 was done by using measurement data from a previous work. Nevertheless the expected behavior of the feature width, namely the decrease with rising dose was verified.

For another measurement the photoresist was applied as image reversal resist (simplified negative tone) and an exposure in a dose range of 15mJ/cm2 - 55mJ/cm2 was investigated. Here the behavior of the increase of the feature width with rising dose could also be confirmed. Nevertheless the top view measurements are not sufficient so further profile measurements were done by breaking the wafer into small pieces and inspect those in side view under the SEM. Here the behaviour is the same but i becomses obvious that the small feature width structures are not fully developed to the bottom of the substrate. The structures with a dose of 25mJ/cm2 exhibits the best quality so an additional series of measurements with this dose value was performed for varying focus positions. It became apparent that for the small structures only the focus range of -2.2ηm evaluable. So the measurement dta was used for the calibration of the image reversal photoresist applied as negative tone resist. Here it is to say, that by using dense line structures the DOF is relative large. Further experiments could be significant and informativ regarding CD and profile behavior for the focus variation by using isolated structures.

The assembled calibration program worked also for the implemented inverted Mack model and was used with the determined CD-SEM measurements values. The behaviour of an increasing feature width with rising dose was confirmed and a best fitness of 0.06ηm was achieved. The comparison of the measurement data with the simulated ones showed that the simulated feature widths were always smaller than the measured ones. The 3D images from the simulations showed a more distinct undercut form of the features than the experimental ones. This explained the samler CD values due to the fact that for the CD-SEM measurements the top CD values had to be used while the simulation CD's were bottom CD values.

At last it is to say that the simulation has to be further developed to improve the prediction of the profile of the structures. The theoretical undercut profiles are not like the real measurements show. The footing effects have also to be taken better into account. Those are apparent for the AZ5214E image reversal photo resist as well as for special negative tone resist for lif-off processes like the AZnLOF2070 photo resist.