Research On Key Technology In Thin-Film Thickness Wideband Monitoring System

The optical properties of optical thin-film are related to each-layer thickness closely, in order to acquire optical thin-film product satisfying the requirement, the thickness must be monitored during the course of deposition. For quarter-wave coating, the turning point monitoring approach of photoelectricity is the most popular which is used for monitoring single wavelength, but transmittance(reflectance) change near the turning point is slow, so the monitoring precision isn't high .With the improvement of complexity and precision requirement of optical thin-film, the film system structure is mostly based on non-quarter-wave type, which need explore the new method for monitoring arbitrary thickness objectively, and requests higher automation level during the course of optical thin-film deposition.Recently one kind of intuitive and easy-used technology, namely thin-film thickness wideband monitoring technology, is getting to mature and has been applied, which is considered as the ideal approach in precisely controlling thin-film wideband characteristics during the process of optical thin-film deposition in the field of thin-film researcher .This method realizes the thin-film monitoring by the way of comparing the measured curve of wideband spectrum scanning with the theoretical target one of calculating according to merit function , and feedbacking the comparative result to the control system. There are many enterprises which developed the relevant product in German, American, and Japan etc. In contrast, the domestic research level is relative lower; the domestic optical coating system can't satisfy the highly precise and complex film system production in automation level.In order to improve the study level of thin-film thickness wideband monitoring system and increase the automation level of domestic optical coating system, the author has systematically studied several key technologies and realization approaches of thin-film thickness wideband monitoring system, which are the online spectrum resolution measurement and the spectrum correction technology in real-time ,the suppression and implementation technology of the background noise, the wavelet de-noising method based on adaptive threshold, the merit function calibration technology based on design and manufacturing integration and so on, and which was sponsored by Education Department and Department of Science & Technology of Shaanxi Provincial Government as key research project. By these researches, the beneficial technology support is expected in improving the study level of thin-film thickness wideband monitoring system and increasing the automation level of domestic optical coating system. Four aspects are discussed emphatically, the following are:1. Beacause the thin-film thickness wideband monitoring system relies on the spectrum curve measured in real-time, online and high quality during the course of thin-film deposition, whose spectral resolution and wavelength position accuracy are key technological parameters affecting on success and failure of monitoring process .Generally the spectral curve measured is obtained by the high-performance spectrometer. Herein firstly the factors are analyzed deeply which affect on the spectral resolution of the grating spectrometer based on CCD detector. It indicates that the contradiction between resolution and luminous flux changes into the one between the sensitivity and the pixel size after the array detector is used, and the corresponding mathematical model is established; Aiming to the thin-film thickness wideband monitoring system studied, whose best slit width is determined by experiment; The non-linear relationship is analyzed which exists in wavelength calibration of the grating spectrometer based on CCD, the corresponding mathematical model is achieved; The corresponding relationship of CCD pixel and the spectrum wavelength is decided in use of the characteristic line of mercury lamp by calibration experiment, the result shows that the cubic polynomial fit can acquire the sub-pixel precision of spectrum wavelength calibration.2. Because CCD of this monitoring system always works on the condition of strong background noise, as well as the stray light produced in vacuum chamber makes the monitoring signal distorted during the process of thin-film deposition, which heavily affects on the correction of spectrum curve measured, even makes the monitoring course failure. For these problems, the signal light modulator controlled by stepper motor is designed to realize the exact match with data acquisition card , the exact sampling in the period of light and dark is completed; The noise signal from background radiation and CCD dark-current noise are effectively suppressed by the signal adjacent subtraction in the period of dark from the signal in the period of light to the same pixel .By experiment, it indicates that the quantitative maximum of noise voltage decreases from 1120 to 200 if the monitoring system begins to work later than 2 to 10 minutes after the system source is switched on ,which ensures the accuracy of the spectrum curve measured.3. For the purpose of suppressing much residual noise of the spectral signal measured after the above processing, which originates mainly from CCD intrinsic noise, on the basis of traditional Donoho wavelet threshold denoising method, wavelet threshold optimization brings forward; Five wavelet decomposition is applied to the noisy signal by use of db4 wavelet, then utilized to process the spectral signal measured. The most appropriate micro-alignment factor g ( m, n )of new threshold function is determined by handling the monitoring signal detected accurately by PMT to reduce true probability of rejection and probability of false ; By experiment, it represents that the wavelet threshold optimization algorithm reserves the signal details perfectly on the basis of smoothing, the peak value error of signal is 0.7%~1.0% and the peak location one is 0.1%~0.3% which satisfies the requirements of thin-film thickness wideband monitoring system. This algorithm is a new pretreatment method for the spectral signal of the monitoring system.4. the optical parameters of thin-film material heavily depend on technological parameters during the course of deposition, inconsistency of the actual parameters with the theoretical ones of thin-film material results in the measured curve serious deviation from the calculated one and multiple extreme value phenomenon of the merit function, this would severely affect on termination judgment, so herein the method puts forward that the target transmittance curve of the layers who will be deposited is corrected on basis of the actual parameters of the layers who have been deposited . This method depends on the spectrum fitting that is to calculate the actual optical parameters of each layer who has been deposited by the spectrum curve measured. When the spectrum curve is fitted, the least square method, the simplex one and the simulated Annealing one are used and their fitting effect is compared by experiment; Lastly thin-film deposition experiment is done in use of layer-by-layer correction technology, the results show that error of thin-film deposition is up to less than 0.4×10-2, the effect is favorable, this technology satisfies the practical requirements.