On-line reoptimization of filter designs for multivariate optical elements.

An automated method for producing multivariate optical element (MOE) interference filters that are robust to errors in the reactive magnetron sputtering process is described. Reactive magnetron sputtering produces films of excellent thickness and uniformity. However, small changes in the thickness of individual layers can have severe adverse effects on the predictive ability of the MOE. Adaptive reoptimization of the filter design during the deposition process can maintain the predictive ability of the final filter by changing the thickness of the undeposited layers to compensate for the errors in deposition. The merit function used, the standard error of calibration, is fundamentally different from the standard spectrum matching. This new merit function allows large changes in the transmission spectrum of the filter to maintain performance.