|
Optical monitoring is a go to method for complex filters deposition; however, it can be easily shown that filter performance is dependent on monitoring strategy. When a non-quarter wave design needs to be deposited, usually one or minimal number of monitoring wavelengths is selected. This allows to use correction algorithms based on swing that are compatible with level-cut monitoring to a great extent. This approach has a significant drawback that it is very difficult to find one or few wavelengths that can be used for all layers of a complex filter. We present a different approach that relies on the selection of the best monitoring wavelength for each layer using pre-defined criteria that secure minimized thickness errors for each individual layer. Wavelength selection process uses several important criteria, such as monitoring wavelength sensitivity to errors in previous layers, transmittance evolution speed versus layer thickness growth, noise of measurement setup… We show that an additional important criteria is spectral resolution of the optical monitoring system and its impact on filter’s spectral response after each layer. Last, to ensure stable deposition we show that some precautions must be made to avoid false turning point detections. Using a binary approach for each criteria (pass or fail), monitoring wavelengths then can be selected automatically based on criteria defined above. In this work we demonstrate that such an approach can be implemented on stable deposition technique such as plasma assisted reactive magnetron sputtering (Bühler HELIOS machine) for different types of filters with various complexities. We illustrate our results for example on an 8-layer beamsplitter, a 37 layer D65 compensation filter, or a 100 layer custom shape spectral filter that are all very sensitive even to small thickness errors. Similar or better spectral performances are achieved compared with classical optical monitoring approaches but with an automatically determined optical monitoring strategy.
|
