|
Optical interference filters allow achieving a wide range of spectral functions. To match with the theoretical performances, fabrication processes require accurate control of each of the layers’ thicknesses and precise refractive index knowledge. However, despite highly stable deposition process, thickness errors tend to accumulate as the number of deposited layers increases, which in turn can lead to severe degradation of the spectral response. Optical monitoring is one the common techniques used to reduce layer thickness errors during deposition. It consists of measuring the filter transmittance, at one or several monitoring wavelengths, and studying its variation during the whole deposition process. Those signals give real-time information on the spectral performances, and in turns, on the deposited optical thickness of the layers. Indeed, for a given filter, the monitoring system can compute the theoretical optical signal during the process and compare it with real-time measurement. Algorithms implemented in this system are then able to at least partially compensate thickness errors from the previous layers, by correcting the expected cut-off points for the current and next layers. In this study, we investigated a specific optical monitoring technique: polychromatic monitoring. This technique uses multiple wavelengths to monitor the layers’ deposition, instead of a unique wavelength for all layers such as conventional monochromatic monitoring. To evaluate the performance of this technique, we designed a numerical model to simulate the online monitoring signals during deposition, based on filter design and refractive index dispersion of the materials. Then, we developed an algorithm that applies several criteria on those simulated signals to automatically determine the most relevant wavelengths to monitor the filter layers - that is to say the ones that are expected to reduce layer thickness error during the deposition process. We also discuss the difficulties and limits when implementing such an approach, such as measured transmission errors or the critical instant for changing optical testglass. Afterwards, we experimentally tested these algorithms of determination of monitoring strategy on an error-sensitive multi-cavity bandpass filter.
|
