Scanning White-Light Interferometry Fingerprints the Polishing Process

Eric Felkel, ZYGO Corp.

Characterizing the effects of polishing optical surfaces beyond a surface roughness parameter may enable superior finishes.

A wide variety of techniques can be used to manufacture high-quality optical surfaces. These range from surface generation, grinding, lapping and full-aperture pitch polishing to more advanced techniques such as deterministic figuring using MRF (magnetorheological finishing), ion beam figuring and computer-controlled polishing. Specialized processes such as diamond turning and polishing are also used for some applications.

Each method removes varying amounts of material, and each leaves behind a signature of its process in the surface. That “fingerprint” can affect the performance of the final optical system in a variety of ways – most typically with less-than-ideal imaging or scattering. Understanding and characterizing the process used to make a surface can be critical to meeting the requirements of a particular application. 3-D optical profilers using scanning white-light interferometry (SWLI) are powerful tools for this characterization.

What is SWLI?

At its core, SWLI profiling uses specialized optical microscope objectives not only to provide the imaging and magnification, but also to measure the 3-D topography of the surface (Figure 1). The microscope’s illuminator projects light through the objective, where a beamsplitter sends some of the light to the reference mirror and some to the part under test. When the optical path length from the beamsplitter to both the reference and test surfaces is equal, the reflected light from both the test and reference surfaces recombines, resulting in interference fringes at the detector. The shape and position of these fringes are directly proportional to the difference in height between the test surface and the reference mirror. The fringe shapes can be thought of as contours of the surface, where the contour intervals are proportional to the illumination wavelength. Processing algorithms further refine this precision to small fractions of the wavelength.

Figure 1. A typical scanning white-light interferometer (SWLI) configuration. An SWLI-based optical profiler easily characterizes surfaces, providing both quantitative texture and qualitative visual information.

To profile the test surface’s topography, the microscope objective is scanned perpendicular to the test surface. A camera and computer system monitor the changing fringe patterns during the scan, and sophisticated algorithms interpret these patterns to construct a 3-D map of the surface.

Using this technique, surface topography measurements at the subnanometer level can be made over any field of view, usually in 5 to 20 s. With a quiet metrology environment, measurement averaging will yield subangstrom measurements using the same equipment. All of this data is obtained completely without contact, eliminating the possibility of damaging a precision sample. In contrast with other microscope-based 3-D topography techniques, SWLI has the distinct advantage that the height resolution of the measurement is consistent across all magnifications, whether the field of view is 20 µm or 20 mm.

Process signatures

Early in the manufacturing process, raw glass is roughly generated and then ground using various sizes of abrasive materials to reach a nominal shape. These surfaces are typically very rough and diffuse. An SWLI profiler highlights the directional nature of the grinding media and unremoved pits and spikes that form as a result of these operations. Minimizing the presence and magnitude of these marks is ideal in keeping the manufacturing time as short as possible. Figure 2 illustrates a typical rough-ground surface manufactured with 320-µm grit. Deep scratches are visible in the dark regions. The crosshatch pattern comes about due to the counter-rotations of the grinding lap and the part.

Figure 2. Ground Zerodur is measured at 10x magnification. Deep scratches are visible in the dark regions.

The polishing methods used to complete a finished optic vary widely in the way that they remove material and in the characteristics of the surface when it is complete. The surface texture of the finished optic can have a dramatic effect on the efficiency and light-scattering characteristics of the final surface, and it is important to know what to expect from these processes. For example, polishing with pitch or a pad can lead to extremely different surface characteristics. In Figures 3a and 3b, these different signatures are clearly visible. The sapphire sample was polished using a pad; the fused silica sample was polished using traditional pitch methods. The pad produces a much more rolling surface with significant low-frequency waviness as compared to the more uniform texture from the pitch process.

The application of the surface is finally what defines whether a particular method is appropriate; the pitch clearly produced a more uniform surface, but pads have advantages of being generally faster and less labor-intensive to set up, maintain and change over.

Sa and rms are not enough

One of the easiest and most common ways of quantifying surface texture of an optic is through a root mean square (Sq) or average roughness (Sa) number. This is a good start, but it cannot tell the entire story of the surface since different methods can produce wildly different surfaces – even if those surfaces have nearly identical Sq or Sa values. This is clearly visible in the surface map images from Figures 3a and 3b, both of which have Sa of 0.22 nm. So how do we distinguish between these surfaces without requiring subjective human analysis? One way is through the power spectral density (PSD) plot.

Figure 3. A surface map and power spectral density (PSD) of pad-polished sapphire (a) and pitch-polished fused silica (b), both shown at 10x magnification. The pad produces a rolling surface with significant low-frequency waviness as compared to the more uniform texture from the pitch process.

Every surface map can be considered a summation of sine waves with varying amplitude. The PSD uses a Fourier analysis to reduce the surface map into these sine waves. The power (the square of the amplitude) of the component waves is plotted as a function of the frequency. This enables quantitative identification of frequency-dependent process signatures that cannot be seen with simpler roughness analysis. The same PSD analysis can be specified across a wide range of spatial frequencies with large-scale form and waviness data from large-aperture figure interferometers, through SWLI profilers, and even down to the finest lateral scale atomic force microscope.

The images in Figures 3a and 3b show two polished surfaces with clearly different surface morphology, measured on a Zygo NewView 7300 SWLI profiler. These two surfaces have the same average roughness value (0.2 nm), but careful analysis of the PSD plots below each shows meaningful differences. The sapphire surface, with its roughness content dominated by lower-frequency waviness, has more energy on the left side of the PSD and then drops off quickly between 10 to 30 cycles per mm. The fused silica, however, is more uniform, with its roughness dropping linearly from 10 to 30 cycles per mm.

Figure 4. Diamond-turned optical surface with PSD. The clear, grooved structure results from the single-point diamond-turning process.

Another example of a surface where PSD is useful is shown in Figure 4 – a diamond-turned optical surface. There is a clear, grooved structure that results from the single-point diamond-turning process. The PSD quickly shows the frequency of this structure as a peak at 45 cycles per mm. A surface such as this is often perfectly acceptable for IR applications, but would be completely unacceptable for most visible and UV applications due to significant scattering and imaging degradation. This is an important point, as the final application ultimately determines which process is viable: A clear understanding of the intended purpose of the optical surface is necessary to judge whether a surface is good or bad, but by using SWLI, roughness parameters and PSD analysis, optics manufacturers have a powerful and flexible set of tools to optimize processes and produce superior performing surfaces.

Meet the author

Eric Felkel is the product manager for noncontact optical profilers at Zygo Corporation in Middlefield, Conn.; email: efelkel@zygo.com.