Accurate metrology improves thin-film filter yield

0501feat2 Logo

Manufacturers of thin-film filters can ensure quality and yield by examining performance factors throughout production. Automated metrology technologies enable accurate, repeatable, and integrated inspections to provide full-process control and continuous improvement.

Matt Novak, Michael Zecchino, and Erik Novak

Thin-film filters play critical roles in developing DWDM devices, allowing higher signal counts per channel, with frequency spacing reduced from many hundreds of gigahertz down to 50 GHz and closer.

Unfortunately, as spacing has decreased, so has filter production yield. The identification of process problems that affect yield is a challenging task. Issues specific to thin-film filter processing include automated metrology technologies that permit accurate process performance characterization. By examining factors throughout the process, critical information can be gained to target process issues, improve production, and ultimately increase filter yield.

To reliably perform systematic process improvement, accurate and stable measurements of filter characteristics must be obtained. Center wavelength (CWL), stop and passbands, ripple, and insertion loss are required for filter assessment, as are reflected power and polarization-dependent loss (PDL). Measuring these parameters early—at the substrate level—serves three purposes: it provides feedback to improve the deposition process; it improves overall production efficiency by ensuring that only passing areas are further processed; and it generates results that can be correlated to downstream measurements to determine the effects of each process on performance.

Since filter requirements vary from application to application, it is important not only to pass or fail each region but also to classify the filter performance. Automated testing at the substrate level lets manufacturers classify and grade filters, so that both top-performing and intermediate-grade parts can be further processed for appropriate applications.

To illustrate how process problems may be identified and characterized, data for a 200-GHz DWDM filter substrate were collected using a substrate mapper.

Coating quality and uniformity depend to a great degree upon the physical geometry of the coating process. Coating parameters change as a function of radial position and, to a lesser extent, angular position on the substrate.

If the substrate is not properly mounted, it may wobble on axis as it is spun during deposition. A substrate coated under such conditions will exhibit symmetrical changes in filter properties. It will be divided essentially in half, with one side passing and one side failing.

A measurement of the sample 200-GHz substrate (see Fig. 1) shows the presence of a nearly straight line of symmetry (from 120° to 330°) about which filter characteristics pass or fail, indicating that the substrate in this coating-run wobbled during deposition. Substrate handling and alignment and/or spindle wear should be reviewed to correct the problem for subsequent deposition runs.

Energy loss through the thin-film filter substrate is due to a combination of scattering and absorption effects. Scattering and absorption may be caused by contamination present in the coatings or by improper substrate preparation and materials.

Total optical-energy loss at individual substrate locations can be calculated by measuring reflection (R) and transmission (T), where

Optical energy loss = 1 - (R + T)

For process development and control it is important to distinguish whether the loss is due to scattering or absorption. For example, substrate roughness may tend to increase scattering, whereas impurities in the films (for example, metal contamination from sputtering of surrounding shield) or poor film stoichiometry will tend to increase absorption.

FIGURE 2. Substrate transmission, reflection, and (1 - ((R + T))) curves show the amount of light lost to scattering and absorption (top). As collection angle increases, so does the width of the stopband because of the transmission of scattered light (bottom).

Thin-film modeling software can be used to examine the effects of various factors on absorption coefficients in the film layers. By using this type of modeling and comparing the model results with actual data (see Fig. 2, top), a diagnosis can be made of various process issues affecting filter performance.

One method for distinguishing scattering is to monitor the width of the stopband at varying apertures. Stopband increases noticeably with scattering (see Fig. 2, bottom), more so for larger apertures. Light that normally would be blocked from transmission through the substrate is now passed at wavelengths lower than the passband limits, thereby widening the stopband.

Comparison of measured transmission-reflection curves with theoretical curves can also be used to help identify control problems in the deposition process. For example, if the 1 - ((R + T)) curve is very close to zero (for example, low scatter and absorption) and the transmission curve deviates significantly from the theoretical curve, then a likely cause may be a layer-thickness control problem.

During process development, distinguishing scattering from absorption and identifying possible deposition-control problems allow researchers to find and correct sources of loss and to establish a process baseline. During subsequent high-volume production, optical energy and stopband width can then be monitored at a constant aperture size that matches the collection angle seen by the filter in the final DWDM component (for example, optical multiplexer/demultiplexer). Changes in these parameters will then indicate a process shift that must be addressed.

Full characterization of the filter manufacturing process requires monitoring of process variables at each step. Performance testing at the filter level indicates the effects of later steps in the filter fabrication process. For example, dicing the substrate into individual filters may alter film stresses, thereby affecting coating performance. By correlating early results from the substrate with testing later in the process, it is possible to determine the effect of these factors on final performance.

Integrated testing methods allow piece-part tracking of filter parameters from substrate to final product. This approach enables correlation of early results to final performance such that predictive performance targets can be established at the substrate level. The Veeco Optium Substrate Mapper and Filter Test and Sort system, for example, employ the same tunable laser diode sources, broad-area detectors, and software analysis, allowing performance tracking throughout the process.

At filter level, thousands of small parts must be handled and tested. Automated handling enables high-speed, high-reliability testing, as well as physical sorting of the parts by selected characteristics (see Fig. 3). For example, filters may be binned according to both CWL and stopband width to identify products that require annealing to adjust the center wavelength.

In addition to performance testing, optical inspection can be employed to reveal chips, cracks, coating defects, contamination, and other errors that may limit the clear aperture of the filter. The output from an optical inspection can reveal heavy edge-chipping and contamination (see Fig. 4). Inspection can be used to classify filters or to act as a pass/fail gate.

An important aspect of inspection is that it can help correlate types and degrees of defects to performance. Combining inspection and performance test data allows process engineers to correlate defects to final performance to distinguish between cosmetic problems and "killer defects." With this knowledge, process designers can set appropriate tolerances that maintain part quality while obtaining highest possible yield.

By accurately assessing the performance of thin-film filters throughout production, manufacturers can establish precise criteria to ensure quality and yield. Highly accurate, repeatable, and integrated inspection, test, and automation are required to provide full-process control and continuous improvement. When the effects of process factors are well-characterized, logical solutions can be readily identified and implemented that directly impact filter-production yield and, ultimately, profitability.

Matt Novak is senior optical engineer, Erik Novak is product manager, and Michael Zecchino is marketing communications manager for the Veeco Metrology Group, 2650 East Elvira Rd., Tucson, AZ 85706. Mike Zecchino can be reached at 520-741-1044 x1022 or

More in Network Test